Systematic review and meta-analysis of the provision of preventive care for modifiable chronic disease risk behaviours by mental health services.

People with mental illness experience increased chronic disease burden, contributed to by a greater prevalence of modifiable chronic disease risk behaviours. Policies recommend mental health services provide preventive care for such risk behaviours. Provision of such care has not previously been synthesised. This review assessed the provision of preventive care for modifiable chronic disease risk behaviours by mental health services. Four databases were searched from 2006 to 2017. Eligible studies were observational quantitative study designs conducted in mental health services, where preventive care was provided to clients for tobacco smoking, harmful alcohol consumption, inadequate nutrition, or inadequate physical activity. Two reviewers independently screened studies, conducted data extraction and critical appraisal. Results were pooled as proportions of clients receiving or clinicians providing preventive care using random effects meta-analyses, by risk behaviour and preventive care element (ask/assess, advise, assist, arrange). Subgroup analyses were conducted by mental health service type (inpatient, outpatient, other/multiple). Narrative synthesis was used where meta-analysis was not possible. Thirty-eight studies were included with 26 amenable to meta-analyses. Analyses revealed that rates of assessment were highest for smoking (78%, 95% confidence interval [CI]:59%-96%) and lowest for nutrition (17%, 95% CI:1%-35%); with variable rates of care provision for all behaviours, care elements, and across service types, with substantial heterogeneity across analyses. Findings indicated suboptimal and variable provision of preventive care for modifiable chronic disease risk behaviours in mental health services, but should be considered with caution due to the very low quality of cumulative evidence. PROSPERO registration: CRD42016049889.

Introduction

People with a mental illness experience greater rates of preventable morbidity, mortality, and a reduced life expectancy up to 30 years compared to people without such an illness (Brown et al., 2010; Laursen et al., 2013; Wahlbeck et al., 2011; Chang et al., 2011; Lawrence et al., 2013; Walker et al., 2015; Olfson et al., 2015; Erlangsen et al., 2017). In high income countries, such disparities are primarily due to a greater prevalence of chronic disease (Brown et al., 2010; Walker et al., 2015; Callaghan et al., 2014; Markkula et al., 2012). This inequitable disease burden is consistently reported to be associated with a greater prevalence of four leading modifiable chronic disease risk behaviours: tobacco smoking, harmful alcohol consumption, inadequate nutrition, and inadequate physical activity (Callaghan et al., 2014; Scott and Happell, 2011; Stanley and Laugharne, 2014; Australian Institute of Health and Welfare, 2012; World Health Organization, 2013).

Care to support and facilitate improvements or reductions in such modifiable chronic disease risk behaviours has been termed ‘preventive care’ (Hensrud, 2000; Woolf et al., 2006; Hulscher et al., 1997). A recommended strategy for addressing such behaviours is the provision of preventive care by health care providers, (Australian Health Ministers' Advisory Council, 2017) with Cochrane systematic review evidence supporting provision of preventive care for such behaviours in general health care settings for risk behaviour improvement (Rees et al., 2013; Hillsdon et al., 2005; Rice et al., 2013; Rigotti et al., 2012; Brunner et al., 2007; Foster et al., 2005). Systematic review evidence demonstrates significant improvements in prevalence of such behaviours among people with a mental illness following the receipt of health promotion programs provided by mental health services (Verhaeghe et al., 2011; Cabassa et al., 2010; Naslund et al., 2017; Happell et al., 2012a). The provision of systematic preventive care is recommended in all health care settings, including mental health services, to all adult clients; (NSW Department of Health, 2017a; NSW Department of Health, 2017b; NSW Mental Health Commission, 2014; National Preventive Health Taskforce, 2008; Royal Australian College of General Practitioners, 2004; Agency for Healthcare Research and Quality, 2010; National Institute for Health and Care Excellence, 2010; National Institute for Health and Care Excellence, 2013; Glasgow et al., 2004; US Department of Health and Human Services, 2009) with tobacco smoking, harmful alcohol consumption, inadequate nutrition, and inadequate physical activity often addressed together in clinical guidelines and recommendations (Royal Australian College of General Practitioners, 2004). The ‘5As’ framework is one recommended approach to facilitate the routine delivery of preventive care (‘ask’ about engagement in risk behaviours; ‘assess’ behaviour risk status and interest in change; ‘advise’ changing/reducing risk behaviours; provide behaviour change ‘assistance’; and ‘arrange’ or refer to other services for behaviour change support (Glasgow et al., 2004).

Despite the effectiveness of preventive care provision in addressing chronic disease risk behaviours in health services, little research has focused on the extent to which any elements of preventive care are delivered by mental health settings. A review of the literature identified two previous reviews of the provision of preventive care within mental health services (Wye et al., 2011; Dickens et al., 2019). A narrative review exploring the provision of care to address smoking within psychiatric inpatient settings cited four studies from the USA and Australia, reported provision of care narratively (Wye et al., 2011). A more recent review (2019) explored mental health nurses' experiences of providing physical health care, with a focus on their attitudes rather than provision of care (Dickens et al., 2019). The review included 41 studies examining general physical health care, sexual health, smoking, physical activity, and nutrition. Six included studies reported advise on regular exercise was provided ‘always or very often’ by 50.4% to 79.7% of participants, advise on healthy diet was provided ‘always or very often’ by 43.4% to 86.7% of participants, and 30.6% to 66.7% of participants reported aiding smoking cessation ‘always or very often’ (Dickens et al., 2019). Whilst the review utilised a broad definition of physical healthcare, limitations included: only exploring mental health nurses' provision of care; outcomes not reported by service type; no meta-analysis due to the broad inclusion criteria and different study methodologies (Dickens et al., 2019). A number of additional individual observational studies have examined preventive care provision for all four risk behaviours in mental health services, utilising varied methodologies (client and clinician self-report, and cross-sectional, pre-post, and interrupted time series surveys; and retrospective medical record audits), however findings have not been quantitatively synthesised previously (Stanley and Laugharne, 2013; Bartlem et al., 2015; Bartlem et al., 2014a; Happell et al., 2013a; Howard and Gamble, 2011). As findings have not been synthesised by mental health service type, it is unknown whether rates of preventive care provision are consistent across service types or otherwise. A systematic synthesis of the extent to which preventive care is provided for all four health risk behaviours to clients of mental health services is lacking. An identification of risk behaviours that receive inadequate care provision, or identifying which elements of care are provided least often could inform the tailoring of future interventions to increase the provision and benefits of preventive care by mental health services.

Given the limitations of previous research, a systematic review and meta-analysis was conducted of the prevalence of preventive care provision for four modifiable chronic disease risk behaviours (tobacco smoking, harmful alcohol consumption, inadequate nutrition, and inadequate physical activity) by mental health services. A secondary aim was to quantify pooled prevalence estimates of preventive care provision for each risk behaviour by care element and service type, and conduct a narrative synthesis where meta-analysis was not possible.

Methods

Review methods and protocol were prospectively registered with PROSPERO [reference number CRD42016049889]. The Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) and Meta-Analysis of Observational Studies in Epidemiology (MOOSE) informed the development of the review protocol (Moher et al., 2015) and reporting of findings (Moher et al., 2009; Stroup et al., 2000).

Search methods

Four databases were searched from January 2006 to August 2017: Medline, PsycINFO, EMBASE, and CINAHL. As clinical practice guidelines for the provision of preventive care have only been introduced recently, (Fiore et al., 2008; Department of Health N, 2009) the search commenced from 2006. Search terms pertained to mental illness and mental health service settings, eligible study designs, four chronic disease risk behaviours of interest, and preventive care practices. Search terms for preventive care practices used a variety of terms to capture care provided by the 5A care elements and any reports of individual care components to account for differences in terminology internationally (Appendix A). Additional sources searched for eligible studies included: first 200 articles from Google Scholar; hand searching of articles published between 2015 and 2017 from Preventive Medicine and Psychiatric Services; and reference lists of included studies. Authors of included studies were contacted and any related publications were requested and screened for eligibility. JBa developed and executed the search strategies with assistance from a research librarian.

Study eligibility criteria

Study type: Descriptive studies including: observational and cross-sectional surveys; retrospective medical record audits; and longitudinal repeated measures were included. Only baseline data from experimental study designs were included.

Setting: Eligible studies were conducted in services with the primary objective of delivering mental health care to adult clients (≥18 years), and could consist of: bed-based (inpatient) mental health services; specialised community (outpatient) mental health services; or community psychosocial support services (non-clinical mental health services). Dual diagnosis services (mental health and substance use treatment) were included if mental health care was a primary care objective. Studies were eligible if preventive care was provided by mental health service staff in the context of routine care delivery.

Outcome measures: Eligible measures included quantitative reporting of the provision or receipt of any elements of preventive care provision for any of four chronic disease risk behaviours (tobacco smoking, harmful alcohol consumption, inadequate nutrition, inadequate physical activity). The World Health Organisation (WHO) has released guidelines for each of the four modifiable risk behaviours, (World Health Organization, 2017; World Health Organization, 2018; World Health Organization, 2011; World Health Organization, 2003) and Australia, (National Health and Medical Research Council, 2009; National Health and Medical Research Council, 2013; Department of Health, 2014; Ministerial Council on Drug Strategy, 2011) and other nations (Piercy et al., 2018; Her Majesty's Government, 2011; Canadian Society for Exercise Physiology, 2011; Ministry of Health, 2015; U.S. Department of Health and Human Services and U.S. Department of Agriculture, 2015; Public Health England, 2016; Health Canada, 2019; Centers for Disease Control and Prevention, 2014; Centers for Disease Control and Prevention, n.d.; Department of Health, 2016; Department of Health and Social Care, 2018; Health Canada, 2018; Ministry of Health, 2007; Butt et al., 2011) similarly have developed ‘country specific’ definitions or guidelines as to what constitutes risk. In the present review, eligible studies were not required to report the definition of risk used for each risk behaviour and variations were expected across countries and studies.

Study selection process

Identified citations were compiled in Endnote, (Anon, 2016) duplicates removed, and remaining citations and abstracts uploaded to Covidence (Anon, n.d.). Two independent reviewers screened titles and abstracts (any of JBa, TB, PW, ES, DT, JD, RC, KB, RH, TR) and relevant full texts (JBa and one of: TB, PW, KB, AM, TR, DT, JD, RH) against predetermined eligibility criteria. Disagreements regarding study eligibility were resolved via consensus or a third reviewer (PW or KB). Corresponding authors were contacted where reported study characteristics were insufficient to determine eligibility, or where outcome data were missing or insufficiently reported.

Data extraction

Data were extracted independently by two authors (JBa and one of: TB, AM, DT, KB, TR) using an Excel-based extraction form, with any inconsistencies resolved by a third reviewer (PW, KB, ES). Where reported, extracted data included study year, setting, sample size, data source (client report, clinician report, medical record audit), study eligibility criteria, clinical and demographic characteristics of mental health service clients and clinicians, clinician engagement in risk behaviours, policies of the mental health service related to preventive care provision, outcome measures, and information required to conduct assessment of methodological quality. Data were extracted separately for each relevant preventive care element and for each risk behaviour (smoking, alcohol, nutrition, physical activity). For intervention studies, only baseline data were extracted. For longitudinal studies, the most recent data were extracted as it was considered to most closely represent current practice.

Data analysis and synthesis

Outcome measures were grouped by data type: clinician reported; client reported; or medical record audit. Preventive care reported in any form/terminology was categorised into the relevant 5A care elements (ask, assess, advise, assist, arrange/refer), and findings were reported using the framework for ease of reporting. The ‘ask’ and ‘assess’ elements of care were combined for analysis. Client reported and audit data were combined for analysis as they represented care provided to individual clients, whereas clinician reported data was reported separately as they represent the proportion of clinicians providing care to an unknown number of clients, consistent with a previous review of smoking care provision in hospitals (Freund et al., 2008).

To facilitate meta-analyses clinician reported data were converted to a common variable of ‘care provision to at least 50% of clients’. For instance, studies reporting care provision to a specific proportion of clients (such as 78%)(Price et al., 2007; Etter et al., 2008; Schacht et al., 2012; Zabeen et al., 2015) were combined with studies reporting care to ‘more than 60% of clients’, (Anderson et al., 2013) ‘50–100%’, (Bartlem et al., 2014a; Chwastiak et al., 2013) and ‘80–100%’ (Bartlem et al., 2014a). Clinician data utilising categorical responses to describe the frequency of care provision (for example, ‘always’) were not pooled in meta-analysis due to the subjective and nonspecific nature. Where a study reported multiple data points for a single care element and behaviour (for instance, 0% of clients were referred to a smoking cessation quitline, and 12% were referred to smoking cessation group education or treatment)(Williams et al., 2015) both the highest and lowest proportions of care provided/received were utilised in separate meta-analyses to reflect the most and least optimistic estimations of care provision. If outcome data were incomplete (missing numerators, denominators, or proportions), data were calculated and backfilled using available data from the studies.

Pre-specified meta-analyses were conducted in RStudio, (RStudio Team, 2015) by JBa with the assistance of a statistician and guidance of ES and RH, where comparable outcome measures were pooled. Random-effects models were utilised to determine pooled prevalence estimates for each preventive care element by each risk behaviour; calculated as proportions and 95% confidence intervals (CIs). Random-effects models were selected as heterogeneity between studies was expected given the different contexts, settings, and delivery of preventive care reported (Deeks et al., 2017; Borenstien et al., 2009). Meta-analyses were conducted where at least two studies contributed data for an outcome measure (Deeks et al., 2017). Heterogeneity in the pooled estimates was assessed via visual inspection of forest plots and consideration of the I2 statistic (Deeks et al., 2017). Where substantial, heterogeneity was explored via subgroup and sensitivity analyses. Where data could not be combined for meta-analysis, narrative summary was undertaken.

Assessment of methodological quality

The methodological quality of each study was assessed independently by two reviewers (JBa and one of: AM, TR, RC, DT, JD, KB) using the Joanna Briggs Institute Critical Appraisal tool: Checklist for Prevalence Studies (Appendix B) (Munn et al., 2015). The following nine domains were assessed: appropriateness of sample frame; appropriateness of participant recruitment; adequacy of sample size; sufficient description of subjects and setting; analysis conducted with sufficient coverage of subgroups; measurement or classification bias; reliable measurement of condition; appropriateness of statistical analysis; and adequacy of response rate. Disagreements were resolved through consensus via discussion or with a third reviewer (KB, JBo).

Subgroup and sensitivity analyses

Where possible, pre-specified subgroup analyses were conducted by: mental health service type: inpatient setting; outpatient setting; and other/multiple service settings. Where heterogeneity was substantial (I2 > 50%; chi-square p < 0.1)(Deeks et al., 2017) post-hoc subgroup analyses were conducted by data type, analysing client reported and medical record audit data separately to explore this. Post-hoc subgroup analyses were also conducted by country.

Pre-specified sensitivity analyses were conducted to exclude studies categorised as high risk of bias (scoring>1 ‘no’ response on methodological quality tool) (Munn et al., 2015). Additional sensitivity analyses were conducted to exclude any studies utilising data that were calculated by the reviewers for the purpose of the review. Furthermore, sensitivity analyses were conducted on clinician reported data to exclude studies reporting any data not in the form of an exact proportion of care provision (for example, ‘80–100%’).

Assessment of confidence in cumulative evidence

Confidence in the cumulative evidence of the primary review outcomes (pre-specified meta-analyses of the provision of care elements by each risk behaviour) were assessed by JBa using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach (GRADE Working Group, 2004; Guyatt et al., 2008a; Guyatt et al., 2008b; Ryan and Hill, n.d.).

Differences between protocol and review

Studies published in languages other than English were originally excluded at full text screening due to inadequate funding for translation as stated in the protocol. However, those studies were subsequently translated to determine eligibility; with consultation from corresponding authors where requested. Data extraction was conducted by two reviewers independently rather than by one reviewer, with a second reviewer checking for consistency, as stated in the protocol. Additionally, the assessment of the cumulative evidence of the main outcomes through the GRADE approach was planned post protocol registration.

Results

Study selection

After duplicates were removed, 16,153 titles and abstracts were screened, 206 studies underwent full-text screening, with 38 studies eligible for inclusion in the review across 48 publications (Fig. 1). Twenty-one studies reported data amenable to meta-analyses; data from 12 studies, which could not be combined for meta-analysis, were summarised narratively; and 5 studies contributed data to both meta-analyses and narrative summaries.

Fig. 1

PRISMA study flow diagram of studies published from 2006 to 2017.

Characteristics of included studies

The majority of the 38 studies were conducted in either Australia (14)(Stanley and Laugharne, 2013; Bartlem et al., 2015; Bartlem et al., 2014a; Happell et al., 2013a; Anderson et al., 2013; Ashton et al., 2010; Siru et al., 2010; Stanton et al., 2015; Stockings et al., 2015; Stockings et al., 2014; Tso et al., 2017; Wye et al., 2010; Wye et al., 2009; Wye et al., 2017) or the USA (11) (Price et al., 2007; Schacht et al., 2012; Chwastiak et al., 2013; Williams et al., 2015; Bolton et al., 2016; Himelhoch et al., 2014; Kilbourne et al., 2011; Leyro et al., 2013; Maki and Bjorklund, 2013; Prochaska et al., 2006; Sarna et al., 2009). Eight described client reported receipt of preventive care, (Bartlem et al., 2015; Etter et al., 2008; Siru et al., 2010; Stockings et al., 2015; Stockings et al., 2014; Leyro et al., 2013; Prochaska et al., 2006; Etter and Etter, 2007) 21 described clinician reported provision of care, (Bartlem et al., 2014a; Happell et al., 2013a; Price et al., 2007; Etter et al., 2008; Schacht et al., 2012; Zabeen et al., 2015; Anderson et al., 2013; Chwastiak et al., 2013; Williams et al., 2015; Ashton et al., 2010; Stanton et al., 2015; Wye et al., 2009; Bolton et al., 2016; Himelhoch et al., 2014; Sarna et al., 2009; Johnson et al., 2009; Haddad et al., 2016; Keizer et al., 2014; Ballbe et al., 2012; Guo et al., 2015; Robson et al., 2013a) and 12 described medical record audit data; (Stanley and Laugharne, 2013; Howard and Gamble, 2011; Williams et al., 2015; Tso et al., 2017; Wye et al., 2010; Wye et al., 2017; Kilbourne et al., 2011; Maki and Bjorklund, 2013; Prochaska et al., 2006; Corradi-Webster et al., 2009; Wu et al., 2013; Parker et al., 2012) where some studies utilised more than one method of data type. Eighteen studies reported data collected in psychiatric inpatient settings; (Howard and Gamble, 2011; Etter et al., 2008; Schacht et al., 2012; Zabeen et al., 2015; Siru et al., 2010; Stanton et al., 2015; Stockings et al., 2015; Stockings et al., 2014; Wye et al., 2010; Wye et al., 2009; Wye et al., 2017; Leyro et al., 2013; Prochaska et al., 2006; Sarna et al., 2009; Etter and Etter, 2007; Haddad et al., 2016; Keizer et al., 2014; Parker et al., 2012) 12 in outpatient settings; (Bartlem et al., 2015; Bartlem et al., 2014a; Price et al., 2007; Anderson et al., 2013; Chwastiak et al., 2013; Tso et al., 2017; Himelhoch et al., 2014; Maki and Bjorklund, 2013; Johnson et al., 2009; Corradi-Webster et al., 2009; Wu et al., 2013; Parker et al., 2012) and 10 in other or multiple mental health service types (Table 1) (Stanley and Laugharne, 2013; Happell et al., 2013a; Williams et al., 2015; Ashton et al., 2010; Bolton et al., 2016; Kilbourne et al., 2011; Ballbe et al., 2012; Guo et al., 2015; Robson et al., 2013a; Parker et al., 2012). Of the 20 studies that utilised client or audit data, 11 reported information on client psychiatric diagnosis; (Bartlem et al., 2015; Howard and Gamble, 2011; Williams et al., 2015; Siru et al., 2010; Stockings et al., 2015; Stockings et al., 2014; Tso et al., 2017; Wye et al., 2017; Leyro et al., 2013; Prochaska et al., 2006; Corradi-Webster et al., 2009) where mood and psychotic disorders were commonly reported (Appendix C). Of the 21 studies that examined clinician reported data, 20 reported clinician profession; (Bartlem et al., 2014a; Happell et al., 2013a; Price et al., 2007; Schacht et al., 2012; Zabeen et al., 2015; Anderson et al., 2013; Chwastiak et al., 2013; Williams et al., 2015; Ashton et al., 2010; Stanton et al., 2015; Wye et al., 2009; Bolton et al., 2016; Himelhoch et al., 2014; Sarna et al., 2009; Johnson et al., 2009; Haddad et al., 2016; Keizer et al., 2014; Ballbe et al., 2012; Guo et al., 2015; Robson et al., 2013a) with clinical/unit managers, and nursing staff frequently surveyed.

Table 1

Included studies reporting the provision of preventive care in mental health services from 2006 to 2017.

Author/year published/country	Year undertaken/setting/sample size (participation rate)	Data source	Chronic disease risk behaviours/preventive care elements*
			Smoking				Alcohol				Nutrition				Physical activity
			As	Ad	At	Ar	As	Ad	At	Ar	As	Ad	At	Ar	As	Ad	At	Ar
Outpatient settings
Andersonb2013Australia	200979 community mental health services79 (94%) service managers	Clinician report	Image 1	Image 2	Image 3	Image 4	Image 5	Image 6	Image 7	Image 8	Image 9	Image 10	Image 11	Image 12	Image 13	Image 14	Image 15	Image 16
Bartlem (Bartlem et al., 2015)2015Australia	2011–201212 community mental health services from one health district in NSW558 (72%) clients	Client report	Image 17	Image 18	Image 19	Image 20	Image 21	Image 22	Image 23	Image 24	Image 25	Image 26	Image 27	Image 28	Image 29	Image 30	Image 31	Image 32
Bartlem (Bartlem et al., 2014a)2014Australia	2010All public community mental health services in one local health district in NSW151 (89%) clinicians	Clinician report	Image 33	Image 34	Image 35	Image 36	Image 37	Image 38	Image 39	Image 40	Image 41	Image 42	Image 43	Image 44	Image 45	Image 46	Image 47	Image 48
Chwastiakb2013USA	20111 community mental health centre154 (71.6%) clinicians	Clinician report	Image 49	Image 50	Image 51	Image 52	Image 53	Image 54	Image 55	Image 56	Image 57	Image 58	Image 59	Image 60	Image 61	Image 62	Image 63	Image 64
Himelhock (Himelhoch et al., 2014)2014USA	20119 community mental health settings95 (100%) clinicians	Clinician report	Image 65	Image 66	Image 67	Image 68	Image 69	Image 70	Image 71	Image 72	Image 73	Image 74	Image 75	Image 76	Image 77	Image 78	Image 79	Image 80
Johnson (Johnson et al., 2009)2009Canada	20068 community mental health teams and 14 contracted community agencies282 (32–38%) clinicians	Clinician report	Image 81	Image 82	Image 83	Image 84	Image 85	Image 86	Image 87	Image 88	Image 89	Image 90	Image 91	Image 92	Image 93	Image 94	Image 95	Image 96
Price (Price et al., 2007)2007USA	200578 community mental health centres with Ohio Department of Mental Health certification80 (53%) psychiatrists	Clinician report	Image 97	Image 98	Image 99	Image 100	Image 101	Image 102	Image 103	Image 104	Image 105	Image 106	Image 107	Image 108	Image 109	Image 110	Image 111	Image 112
Corradi-Webster (Corradi-Webster et al., 2009)2009Brazil	2000–20041 psychiatric outpatient clinic127 records	Medical record audit	Image 113	Image 114	Image 115	Image 116	Image 117	Image 118	Image 119	Image 120	Image 121	Image 122	Image 123	Image 124	Image 125	Image 126	Image 127	Image 128
Makib2013USA	nr1 community mental health centre129 records	Medical record audit	Image 129	Image 130	Image 131	Image 132	Image 133	Image 134	Image 135	Image 136	Image 137	Image 138	Image 139	Image 140	Image 141	Image 142	Image 143	Image 144
Tso (Tso et al., 2017)2017Australia	2014–2015Community mental health clinics at 2 public hospitals251 records	Medical record audit	Image 145	Image 146	Image 147	Image 148	Image 149	Image 150	Image 151	Image 152	Image 153	Image 154	Image 155	Image 156	Image 157	Image 158	Image 159	Image 160
Wu (Wu et al., 2013)2013UK	2008–2011South London and Maudsley (SLaM) Case Register5588 records	Medical record audit	Image 161	Image 162	Image 163	Image 164	Image 165	Image 166	Image 167	Image 168	Image 169	Image 170	Image 171	Image 172	Image 173	Image 174	Image 175	Image 176
Inpatient settings
Etterb2008Switzerland	2003–20062 inpatient psychiatric units within a hospital2006–77 (67.5%) patients	Client report	Image 177	Image 178	Image 179	Image 180	Image 181	Image 182	Image 183	Image 184	Image 185	Image 186	Image 187	Image 188	Image 189	Image 190	Image 191	Image 192
	2005–20062 inpatient psychiatric units within a hospital2006–57 (91.9%) clinicians	Clinician report	Image 193	Image 194	Image 195	Image 196	Image 197	Image 198	Image 199	Image 200	Image 201	Image 202	Image 203	Image 204	Image 205	Image 206	Image 207	Image 208
Etterb2007Switzerland	2003–20042 inpatient psychiatric units within a hospital49 (86%) patients	Client report	Image 209	Image 210	Image 211	Image 212	Image 213	Image 214	Image 215	Image 216	Image 217	Image 218	Image 219	Image 220	Image 221	Image 222	Image 223	Image 224
Leyro (Leyro et al., 2013)2013USA	2006–20102 psychiatric hospitals324 (71% and 79% per hospital) patients	Client report	Image 225	Image 226	Image 227	Image 228	Image 229	Image 230	Image 231	Image 232	Image 233	Image 234	Image 235	Image 236	Image 237	Image 238	Image 239	Image 240
Prochaska (Prochaska et al., 2006)2006USA	nr1univeristy-based inpatient psychiatry unit100 (87%) patients	Client report	Image 241	Image 242	Image 243	Image 244	Image 245	Image 246	Image 247	Image 248	Image 249	Image 250	Image 251	Image 252	Image 253	Image 254	Image 255	Image 256
	100 records	Medical record audit	Image 257	Image 258	Image 259	Image 260	Image 261	Image 262	Image 263	Image 264	Image 265	Image 266	Image 267	Image 268	Image 269	Image 270	Image 271	Image 272
Siru (Siru et al., 2010)2010Australia	2008Department of Psychiatry in a major teaching hospital64 (nr) patients	Client report	Image 273	Image 274	Image 275	Image 276	Image 277	Image 278	Image 279	Image 280	Image 281	Image 282	Image 283	Image 284	Image 285	Image 286	Image 287	Image 288
Stockings (Stockings et al., 2014)2014Australia	2010–20113 psychiatric inpatient units in a large, regional public hospital205 (69%) patients	Client report	Image 289	Image 290	Image 291	Image 292	Image 293	Image 294	Image 295	Image 296	Image 297	Image 298	Image 299	Image 300	Image 301	Image 302	Image 303	Image 304
Stockings (Stockings et al., 2015)2015Australia	2009–20103 psychiatric inpatient units in a large, regional public hospital181 (90.9%) patients	Client report	Image 305	Image 306	Image 307	Image 308	Image 309	Image 310	Image 311	Image 312	Image 313	Image 314	Image 315	Image 316	Image 317	Image 318	Image 319	Image 320
Haddad (Haddad et al., 2016)2016UK	nr1 low secure forensic psychiatric inpatient unit57 (90.5%) clinicians	Clinician report	Image 321	Image 322	Image 323	Image 324	Image 325	Image 326	Image 327	Image 328	Image 329	Image 330	Image 331	Image 332	Image 333	Image 334	Image 335	Image 336
Keizer (Keizer et al., 2014)2014Switzerland	2009Department of Mental Health and Psychiatry in a large hospital155 (72.4%) clinicians	Clinician report	Image 337	Image 338	Image 339	Image 340	Image 341	Image 342	Image 343	Image 344	Image 345	Image 346	Image 347	Image 348	Image 349	Image 350	Image 351	Image 352
Sarna (Sarna et al., 2009)2009USA	nrAdult psychiatric inpatient settings in a Magnet-designated health care facility100 (100%) nurses	Clinician report	Image 353	Image 354	Image 355	Image 356	Image 357	Image 358	Image 359	Image 360	Image 361	Image 362	Image 363	Image 364	Image 365	Image 366	Image 367	Image 368
Schacht (Schacht et al., 2012)2012USA	2011206 state inpatient psychiatric facilities165 (80%) facility directors	Clinician report	Image 369	Image 370	Image 371	Image 372	Image 373	Image 374	Image 375	Image 376	Image 377	Image 378	Image 379	Image 380	Image 381	Image 382	Image 383	Image 384
Stanton (Stanton et al., 2015)2015Australia	nrinpatient psychiatric facilities in a regional city in QLD34 (nr) nurses	Clinician report	Image 385	Image 386	Image 387	Image 388	Image 389	Image 390	Image 391	Image 392	Image 393	Image 394	Image 395	Image 396	Image 397	Image 398	Image 399	Image 400
Wye (Wye et al., 2009)2009Australia	2006All publicly funded psychiatric inpatient units in NSW123 (94%) nurse unit managers	Clinician report	Image 401	Image 402	Image 403	Image 404	Image 405	Image 406	Image 407	Image 408	Image 409	Image 410	Image 411	Image 412	Image 413	Image 414	Image 415	Image 416
Zabeenb2015UK	2010Random sample of inpatient psychiatric units across England147 (67%) unit managers	Clinician report	Image 417	Image 418	Image 419	Image 420	Image 421	Image 422	Image 423	Image 424	Image 425	Image 426	Image 427	Image 428	Image 429	Image 430	Image 431	Image 432
Howard (Howard and Gamble, 2011)2011UK	nr2 acute wards from a large mental health trust28 records	Medical record audit	Image 433	Image 434	Image 435	Image 436	Image 437	Image 438	Image 439	Image 440	Image 441	Image 442	Image 443	Image 444	Image 445	Image 446	Image 447	Image 448
Wye (Wye et al., 2010)2010Australia	2005–20061 large adult psychiatric hospital1000 (99%) records	Medical record audit	Image 449	Image 450	Image 451	Image 452	Image 453	Image 454	Image 455	Image 456	Image 457	Image 458	Image 459	Image 460	Image 461	Image 462	Image 463	Image 464
Wye (Wye et al., 2017)2017Australia	2009–20102 general locked adult inpatient psychiatric facilities in one health district in NSW1054 records	Medical record audit	Image 465	Image 466	Image 467	Image 468	Image 469	Image 470	Image 471	Image 472	Image 473	Image 474	Image 475	Image 476	Image 477	Image 478	Image 479	Image 480
Other settings
Ashton (Ashton et al., 2010)2010Australia	200745 government and non-government mental health organisations in Adelaide, SA324 (60%) team members	Clinician report	Image 481	Image 482	Image 483	Image 484	Image 485	Image 486	Image 487	Image 488	Image 489	Image 490	Image 491	Image 492	Image 493	Image 494	Image 495	Image 496
Ballbe (Ballbe et al., 2012)2012Spain	2008–2009186 Inpatient and outpatient mental health services in Catalonia186 (96.9%) Clinical managers	Clinician report	Image 497	Image 498	Image 499	Image 500	Image 501	Image 502	Image 503	Image 504	Image 505	Image 506	Image 507	Image 508	Image 509	Image 510	Image 511	Image 512
Boltona2016USA	2015Members of the American Psychiatric Nurses Association26 (nr) clinicians	Clinician report	Image 513	Image 514	Image 515	Image 516	Image 517	Image 518	Image 519	Image 520	Image 521	Image 522	Image 523	Image 524	Image 525	Image 526	Image 527	Image 528
Guo (Guo et al., 2015)2015Taiwan	nr2 community psychiatric hospitals providing inpatient and outpatient services193 (96.9%) clinicians	Clinician report	Image 529	Image 530	Image 531	Image 532	Image 533	Image 534	Image 535	Image 536	Image 537	Image 538	Image 539	Image 540	Image 541	Image 542	Image 543	Image 544
Happella2013Australia	2012Members of the Australian College of Mental Health Nurses559 (19.6%) mental health nurses	Clinician report	Image 545	Image 546	Image 547	Image 548	Image 549	Image 550	Image 551	Image 552	Image 553	Image 554	Image 555	Image 556	Image 557	Image 558	Image 559	Image 560
Robson (Robson et al., 2013a)2013UK	2006–2007Mental health nurses recruited from a large National Health Service Mental Health Trust in the UK585 (52%) mental health nurses	Clinician report	Image 561	Image 562	Image 563	Image 564	Image 565	Image 566	Image 567	Image 568	Image 569	Image 570	Image 571	Image 572	Image 573	Image 574	Image 575	Image 576
Williams (Williams et al., 2015)2015USA	201230 outpatient or partial-hospitalisation settings of state-wide behavioural health agency18 (90%) clinicians	Clinician report	Image 577	Image 578	Image 579	Image 580	Image 581	Image 582	Image 583	Image 584	Image 585	Image 586	Image 587	Image 588	Image 589	Image 590	Image 591	Image 592
	100 records	Medical record audit	Image 593	Image 594	Image 595	Image 596	Image 597	Image 598	Image 599	Image 600	Image 601	Image 602	Image 603	Image 604	Image 605	Image 606	Image 607	Image 608
Kilbourne (Kilbourne et al., 2011)2011USA	2006–2007VA Mental Health Programs with and without colocated general medical services7514 (7.1%) records	Medical record audit	Image 609	Image 610	Image 611	Image 612	Image 613	Image 614	Image 615	Image 616	Image 617	Image 618	Image 619	Image 620	Image 621	Image 622	Image 623	Image 624
Parker (Parker et al., 2012)2012UK	2010–2011Adult mental health treatment services in the United Kingdom's largest Mental Health Trust85 inpatient records2028 community patient records	Medical record audit	Image 625	Image 626	Image 627	Image 628	Image 629	Image 630	Image 631	Image 632	Image 633	Image 634	Image 635	Image 636	Image 637	Image 638	Image 639	Image 640
Stanley (Stanley and Laugharne, 2013)2013Australia	2011–20121 Fremantle adult psychiatric hospital with inpatient and outpatient services, 1 Kimberley rural mental health and drug service56 Kimberley records228 Fremantle records	Medical record audit	Image 641	Image 642	Image 643	Image 644	Image 645	Image 646	Image 647	Image 648	Image 649	Image 650	Image 651	Image 652	Image 653	Image 654	Image 655	Image 656
Total Image 657c			23/41	22/41	25/41	12/41	7/41	3/41	1/41	2/41	4/41	8/41	0/41	2/41	4/41	9/41	0/41	4/41

*As- Ask/ Assess, Ad- Advise, At- Assist, Ar- Arrange.

- Assessed.

- Not assessed.

nr = Not reported.

Data not reported in paper, analysed for review purposes.

Study reported incomplete outcome data where either the numerator or denominator of care provision/receipt was not reported and calculations were assumed.

Total number of data sources (Stanley and Laugharne, 2013) from 38 studies reporting on the provision of preventive care.

Provision of preventive care

Most (36) studies reported the receipt/provision of smoking cessation care, with fewer reporting the receipt/provision of care for: inadequate physical activity (10); inadequate nutrition (9); and harmful alcohol consumption (8; Table 1).

With regard to elements of care provision: the proportions of clients reporting being asked/clinicians reporting ‘asking’ about and/or ‘assessing’ extent of client risk behaviour was reported in 23 studies; ‘advising’ the modification of risk behaviours was reported in 25 studies; ‘assisting’ with behaviour change was reported in 22 studies; and ‘arranging’ referral or further behaviour change support was reported in 13 studies (Table 1).

Most studies were of low methodological quality; 25/38 classified as high risk of bias. Most studies adequately described the subjects and setting, and had adequate response rates. Just two studies utilised appropriate statistical analysis, and no studies utilised a valid measurement of care provision; nor did data analysis provide sufficient coverage of subgroups-that is, coverage bias resulting from differing response rates among subgroups (where applicable; Appendix D).

Meta-analysis of prevalence of preventive care provision

Due to space constraints, meta-analysis results presented below reflect the highest estimations of care provision across clinician, client and audit reported data. Refer to Table 2 for results of the lowest estimations, and Appendix E for all Forest plots.

Table 2

Summary of effects of all meta-analyses of included studies published from 2006 to 2017.

Outcomea	% (95% CI)	I2 (%)	p	n	n of studies
Meta-analysis results:
Overall clinician reported data- highest estimates of care provision
Ask/assess
Smoking	78 (59%–96%)	97.7	<0.01	515	4
Advise
Smoking	46 (31%–61%)	88.5	<0.01	384	3
Nutrition	54 (48%–59%)	0	0.78	305	2
Physical activity	72 (49%–95%)	95.4	<0.01	304	2
Assist
Smoking	52 (31%–73%)	94.2	<0.01	339	4
Arrange
Smoking	30 (1%–59%)	95.6	<0.01	211	2
Overall clinician reported data- lowest estimates of care provision
Ask/Assess
Smoking	78 (59%–96%)	97.7	<0.01	515	4
Advise
Smoking	43 (32%–54%)	78.5	<0.01	384	3
Nutrition	54 (48%–59%)	0	0.78	305	2
Physical activity	72 (49%–95%)	95.4	<0.01	304	2
Assist
Smoking	39 (12%–67%)	97.8	<0.01	339	4
Arrange
Smokingb	7 (0%–20%)	90.0	<0.01	229	2
Overall client and audit reported data- highest estimates of care provision
Ask/assess
Smoking	54 (38%–71%)	99.7	<0.01	10,574	12
Alcohol	62 (42%–81%)	99.0	<0.01	3240	6
Nutritionb	17 (0%–35%)	98.6	<0.01	813	3
Physical activityb	35 (0%–72%)	99.6	<0.01	641	3
Advise
Smoking	28 (14%–42%)	98.4	<0.01	1880	10
Alcoholb	42 (0%–100%)	98.8	<0.01	228	2
Nutrition	47 (5%–90%)	95.6	<0.01	152	2
Physical activityb	46 (0%–100%)	99.5	<0.01	190	2
Assist
Smoking	37 (13%–61%)	99.7	<0.01	3141	8
Arrange
Smokingb	21 (0%–49%)	98.9	<0.01	388	3
Physical activity	35 (10%–59%)	87.8	<0.01	190	2
Overall client and audit reported data- lowest estimates of care provision
Ask/Assess
Smoking	41 (22%–61%)	99.6	<0.01	10,574	12
Alcohol	62 (42%–81%)	99.0	<0.01	3240	6
Nutritionb	17 (0%–35%)	98.6	<0.01	813	3
Physical activityb	35 (0%–72%)	99.6	<0.01	641	3
Advise
Smoking	25 (10%–40%)	99.2	<0.01	1880	10
Alcoholb	42 (0%–100%)	98.8	<0.01	228	2
Nutrition	47 (5%–90%)	95.6	<0.01	152	2
Physical activityb	46 (0%–100%)	99.5	<0.01	190	2
Assist
Smoking	31 (7%–56%)	99.8	<0.01	3141	8
Arrange
Smokingb	3 (0%–7%)	85.5	<0.01	388	3
Physical activityb	10 (0%–29%)	84.0	0.01	190	2

Meta-analyses were not possible for all health behaviours by all care elements due to insufficient numbers of studies (n < 2) contributing data.

In some cases approximate confidence intervals for the proportion gave limits outside 0 and 1. These cases have been truncated to 0 or 1 as appropriate.

Ask/Assess: Client and audit reported care data were available for meta-analysis for all four behaviours, however clinician reported data were only available for smoking. According to the combined client and audit reported data, assessment was most likely to occur for alcohol use (62%, 95% CI: 42%–81%; I2 99%; participants = 3240; studies = 6), followed by smoking (54%, 95% CI: 38%–71%; I2 99.7%; participants = 10,574; studies = 12; Fig. 2), physical activity (35%, 95% CI: −1%-72%; I2 99.6%; participants = 641; studies = 3), and nutrition (17%, 95% CI: 1%–35%; I2 98.6%; participants = 813; studies = 3). Assessment of smoking via clinician report was somewhat higher (78%, 95% CI: 59%–96%; I2 97.7%; participants = 515; studies = 4). Heterogeneity in pooled estimates was substantial (I2 > 50%; Appendices E, F).

Fig. 2

Forest plots of the overall client and audit reported highest estimates of smoking cessation care provision by care element: a) ask/assess, b) advise, c) assist, and d) arrange.

Note: Error-bars represent 95% confidence intervals; included studies published from 2006 to 2017. At times, 95% confidence intervals exceed the boundaries of the proportion (between 0 and 1), as the random effects binomial proportion interval (Wald interval) assumes a normal distribution.

Advise: Client and audit reported data for “advice” were available for meta-analyses for all behaviours, and clinician reported data for all behaviours except alcohol. The receipt of client and audit reported advice to change at risk behaviours was similar for nutrition (47%, 95% CI: 5%–90%; I2 95.6%; participants = 152; studies = 2), physical activity (46%, 95% CI: −30%-123%; I2 99.5%; participants = 190; studies = 2), and alcohol (42%, 95% CI: −19%-102%; I2 98.8%; participants = 228; studies = 2), yet lower for smoking (28%, 95% CI: 14%–42%; I2 98.4%; participants = 1880; studies = 10; Fig. 2). Similar proportions of advice were provided via clinician report for smoking (46%, 95% CI: 31%–61%; I2 88.5%; participants = 384; studies = 3) and nutrition (54%, 95% CI: 48%–59%; I2 0%; participants = 305; studies = 2), with advice for physical activity somewhat higher (72%, 95% CI: 49%–95%; I2 95.4%; participants = 211; studies = 2). Heterogeneity was substantial, with the exception of clinician reported nutrition advice; however this value is a probable underestimation of heterogeneity due to the small number of data points (Appendices E, F) (von Hippel, 2015).

Assist: Meta-analyses were conducted for clinician and client/audit reported data and were possible for the provision of assistance for smoking only. The pooled prevalence of clinician reported provision of assistance for smoking was 52% (95% CI: 31%–73%; I2 94.2%; participants = 339; studies = 4), whilst client and audit reported receipt was 37% (95% CI 13%–61%; I2 99.7%; participants = 3141; studies = 8; Fig. 2). Heterogeneity was substantial (Appendices E, F).

Arrange: Client and audit reported data enabled meta-analysis for smoking and physical activity outcomes, whilst clinician reported data were available for meta-analysis of smoking only. The pooled prevalence of clinician reported arrangement of smoking cessation care was 30% (95% CI: 1%–59%; I2 95.6%; participants = 211; studies = 2), whilst the client and audit reported receipt was 21% (95% CI: −7%-49%; I2 98.9%; participants = 388; studies = 3; Fig. 2). Client and audit reported receipt of care arranged for physical activity was 35% (95% CI: 10%–59%; I 87.8%; participants = 190; studies = 2). Heterogeneity was substantial (Appendices E, F).

GRADE ratings for outcome measures were initially classified as ‘low’ as the included data were observational (Guyatt et al., 2008b). The quality of evidence (GRADE) in all meta-analyses for all elements of care provision, across all four risk behaviours, was downgraded to ‘very low’ due to risk of bias, and/or inconsistency, and/or imprecision of results (Appendix F) (Ryan and Hill, n.d.).

Narrative synthesis

Seventeen studies reported data not conducive to meta-analysis. Thirteen of these studies reported data using categorical responses describing the frequency of clinician care provision (for example, ‘always’) (Happell et al., 2013a; Williams et al., 2015; Ashton et al., 2010; Stanton et al., 2015; Wye et al., 2009; Bolton et al., 2016; Himelhoch et al., 2014; Sarna et al., 2009; Johnson et al., 2009; Haddad et al., 2016; Keizer et al., 2014; Ballbe et al., 2012; Robson et al., 2013a). Two of the 17 studies, one utilising client reported data and one clinician report, were not included in the meta-analyses as no other studies reported data points for the same risk behaviour and care element (Bartlem et al., 2015; Bartlem et al., 2014a). Five of the 17 studies, also utilised categorical responses, reporting preventive care outcomes in mean scores rather than proportions: for smoking: ask/assess; (Price et al., 2007; Ashton et al., 2010) advise; (Price et al., 2007) assist; (Happell et al., 2013a; Price et al., 2007; Bolton et al., 2016; Guo et al., 2015) and arrange; (Price et al., 2007) and advice for nutrition and physical activity; (Bolton et al., 2016) and assistance for alcohol, nutrition, and physical activity (Appendix C) (Happell et al., 2013a).

Ask/Assess: The proportions of clinicians reporting ‘often’ or ‘always’ providing assessment care for smoking was 26.1–100% (Williams et al., 2015; Ashton et al., 2010; Himelhoch et al., 2014; Sarna et al., 2009; Johnson et al., 2009; Ballbe et al., 2012). Proportions of clinicians reporting providing 80–100% of clients with assessment was greatest for alcohol (89.4%), followed by physical activity (59.6%), and nutrition (13.2%) (Bartlem et al., 2014a).

Advise: The highest proportions of clinicians reporting ‘often’ or ‘always’ providing advice were found for nutrition (61–100%)(Happell et al., 2013a; Bolton et al., 2016; Haddad et al., 2016; Robson et al., 2013a) and physical activity (53–100%), (Happell et al., 2013a; Stanton et al., 2015; Bolton et al., 2016; Haddad et al., 2016; Robson et al., 2013a) with a more variable range for smoking (16–100%) (Williams et al., 2015; Wye et al., 2009; Himelhoch et al., 2014; Sarna et al., 2009; Johnson et al., 2009). Additionally, 80.1% of clinicians were reported to provide advice to reduce alcohol consumption to 80–100% of clients (Bartlem et al., 2014a).

Assist: Between 5 and 92.9% of clinicians reported ‘often’ or ‘always’ providing clients with assistance for smoking, (Happell et al., 2013a; Williams et al., 2015; Wye et al., 2009; Bolton et al., 2016; Himelhoch et al., 2014; Sarna et al., 2009; Johnson et al., 2009; Haddad et al., 2016; Ballbe et al., 2012; Robson et al., 2013a) and 86.2% reported ‘often’ or ‘always’ providing assistance for alcohol consumption (Happell et al., 2013a). No studies reported on assistance for nutrition or physical activity.

Arrange: Between 8 and 94% of clinicians reported ‘often’ or ‘always’ arranging care for smoking for clients (Williams et al., 2015; Wye et al., 2009; Himelhoch et al., 2014; Sarna et al., 2009; Ballbe et al., 2012). Higher proportions of clinicians reported arranging care for 80–100% of clients for alcohol consumption (60.9%), compared to physical activity (40.1%) and nutrition (22.5%) (Bartlem et al., 2014a). Additionally, 38% of clients reported receiving care for alcohol, and 43% reported receiving care for nutrition (Appendix C) (Bartlem et al., 2015).

Service type: Pre-specified subgroup analyses were conducted by mental health service type: inpatient; outpatient; and other/multiple settings (Appendix E, F). With respect to subgroup analyses of clinician reported data, provision of care did not differ significantly in inpatient or outpatient settings relative to the overall pooled estimate.

Subgroup analyses of client and audit reported data of care provision by setting revealed non-significant trends of variability in pooled prevalence estimates between settings. Client and audit reported care for smoking ask/assessment was lower in inpatient settings (31%), and higher in outpatient (70%) and other settings (68%) relative to the overall pooled estimate (54%). Ask/assessment of alcohol consumption was lower in outpatient settings (49%) and higher in other settings (83%) compared to the pooled estimate (62%). Additionally, receipt for smoking cessation assistance was lower in other settings (23%) and higher in inpatient settings (45%) relative to the overall polled estimate (37%; Appendix F).

Heterogeneity remained substantial for all individual subgroup analyses with the exception of: clinician reported provision of nutrition advice in outpatient settings (I2 0%); and client receipt of alcohol ask/assessment in other settings (I2 0%; Appendix F), however these values are likely an underestimation of heterogeneity due to small number of data points (von Hippel, 2015). To explore substantial heterogeneity, a further post-hoc subgroup analysis of client and audit data separately by setting type was conducted, however heterogeneity remained substantial with the exception of audit data of smoking advice in inpatient settings (I2 0%; Appendix F).

Post-hoc subgroup analyses were conducted by country, where possible (Australia only for clinician reported data; and Australia, UK, and US for client and audit data). With respect to Australian clinician reported data, no significant differences were found compared to the overall meta-analyses. Moreover, subgroup analyses using client and audit reported data yielded no significant differences between the countries or the overall estimates (Appendix F).

Sensitivity analyses: Pre-specified sensitivity analyses were conducted to exclude studies at high risk of bias. All studies reporting clinician data were assessed at high risk of bias; therefore analysis was not possible. Sensitivity analyses of client and audit reported care excluding studies at high risk of bias revealed a similar pooled prevalence to all possible comparisons (within 6%) with the overall meta-analyses results (Appendix F).

Post-hoc sensitivity analyses were conducted to exclude any data calculated by the authors for the purpose of the review. This eliminated all clinician reported analyses bar one, where clinician reported ask/assess of smoking was similar to overall care provision estimates; as were all client and audit reported sensitivity analyses excluding calculated data (Appendix F).

Finally, sensitivity analyses utilising only exact proportions of clinician reported care provision estimates yielded just one analysis, where the pooled prevalence of provision of assistance for smoking did not differ from the overall estimate (Appendix F).

Discussion

This is the first review to comprehensively synthesise the international evidence on the provision of preventive care by mental health services for four modifiable chronic disease risk behaviours by individual preventive care elements and by service type. Meta-analysis revealed sub-optimal levels of care provision (defined as <80% of clients in receipt of care in previous research)(Bartlem et al., 2014b; McElwaine et al., 2014; Freund et al., 2005) across clinician and client/audit reported data for each of the four risk behaviours and all analysed care elements. Relatively few of the included studies examined the provision of care for behaviours other than smoking and across all care elements. Estimations of care provision by care element varied across risk behaviours revealing no clear patterns. Similarly, subgroup analyses did not reveal any consistent trends across settings or country, likely due to the small number of studies available. Further research is needed to explore the extent to which different mental health settings provide preventive care for harmful alcohol consumption, inadequate nutrition, and inadequate physical activity.

One of the key findings of the review was the wide variation in how care was measured across the included studies; hampering comparisons across studies. Moreover, the creation of the pooled clinician measure lacked specificity whereby estimates of care could only be calculated for ‘at least 50% of clients’. Despite these constraints, pooled estimates of care provision suggest between one-third and three-quarters (36–78%) of clinicians are providing at least 50% of clients with preventive care for the behaviours and care elements analysed. Whilst these findings suggest overall sub-optimal levels of care provision, many clinicians are currently providing care to at least 50% of clients suggesting an attempt to incorporate preventive care into service practice where clinicians are aware of the requirement (NSW Department of Health, 2017a; NSW Department of Health, 2017b; NSW Mental Health Commission, 2014; National Preventive Health Taskforce, 2008; Agency for Healthcare Research and Quality, 2010; National Institute for Health and Care Excellence, 2010; National Institute for Health and Care Excellence, 2013; US Department of Health and Human Services, 2009). However, less than a third of studies reported any details on any policies or procedures operational within the settings studied regarding the requirements of preventive care provision (Stanley and Laugharne, 2013; Bartlem et al., 2015; Bartlem et al., 2014a; Stockings et al., 2015; Stockings et al., 2014; Wye et al., 2010; Leyro et al., 2013; Etter and Etter, 2007; Keizer et al., 2014; Parker et al., 2012).

Another key finding of the review was the dearth of previous research assessing the extent to which care is provided for behaviours other than smoking in mental health services. No included studies reported on the provision of assistance for improving nutrition or physical activity by clinician nor client, nor audit reported data. Similarly no studies reported provision of assistance for alcohol consumption by client or audit report, and only one study (in the narrative synthesis report)(Happell et al., 2013a) reported on the provision of assistance for alcohol consumption by clinician reported data. As such, the current provision of care in the form of assistance to address nutrition, physical activity, and alcohol consumption is largely unknown and requires further investigation in subsequent research.

The findings of this review confirm sub-optimal reports of the provision of preventive care reported in individual identified studies and the previous narrative review of smoking cessation care provision in inpatient settings; (Wye et al., 2011) suggesting a need for further research to address barriers to the provision of preventive care. Previous research conducted in mental health settings has identified multiple barriers to the provision preventive care at the clinician and service level (Price et al., 2007; Anderson et al., 2013; Chwastiak et al., 2013; Ashton et al., 2010; Johnson et al., 2009; Johnson and Fry, 2013; Robson et al., 2013b; Happell et al., 2012b; Hyland et al., 2003; Organ et al., 2010; Happell et al., 2013b; Dunbar et al., 2010; Nash, 2005). To address such barriers, various strategies have been tested and demonstrated to reduce clinician burden in providing preventive care and increase the provision of care such as: reduction of the ‘5As’ model to ‘2As and R’; (Schroeder, 2005; Revell and Schroeder, 2005) incorporation of tools such as prompts, recording and arrange/referral protocols; (Shojania et al., 2009; Wolfenden et al., 2009; Krist et al., 2008) and training in the provision of care and referral options (Sheffer et al., 2012). Alternatively, the institution of a specialist preventive care provider embedded in mental health services, relieving mental health clinicians of the role of preventive care provision, has been trialled through limited research and found to increase care provision (Osborn et al., 2010; McKenna et al., 2014; Cunningham et al., 2013). Further research could focus on the design of practice change interventions to address identified barriers and increase the provision of care across mental health services. Additionally, the review investigated care provision for the four leading modifiable chronic disease risk behaviours. Future research could examine the provision of care by mental health services for other modifiable risk behaviours, such as sleep; more recently recognised as a risk factor for chronic disease (von Ruesten et al., 2012; Liu et al., 2013).

Study limitations and strengths

This is the first review to comprehensively synthesise the international published evidence on the provision of preventive care by mental health services for four health risk behaviours by individual preventive care elements and by service type.

The review is limited by the variability and heterogeneity of included individual studies. Substantial heterogeneity remained throughout the majority of meta-analyses and subgroup analyses, contributing to the very low quality assessment of the cumulative evidence. As included studies were observational to reflect real-world practice and not conducted under strict controlled conditions, some heterogeneity is to be expected. Such heterogeneity might be explained by between-study differences in methodology including measurements and definitions of care elements used to assess care provision, or by population characteristics such as country; clinician qualification/profession; or psychiatric diagnosis of clients (; Glasziou and Sanders, 2002). Heterogeneity may also result from true differences in provision of care impacted by unknown characteristics, such as whether or not the settings studied operated under specific preventive care policies; as changes in smoke-free policies have resulted in changes in smoking cessation care provision (Etter et al., 2008). As mentioned above, limited studies provided information on whether or not a service operated under specific preventive care policies, and few studies that mentioned active policies or procedures provided any description of such procedures (Stanley and Laugharne, 2013; Bartlem et al., 2015; Bartlem et al., 2014a; Stockings et al., 2015; Stockings et al., 2014; Keizer et al., 2014). Future research on preventive care provision in mental health settings could seek to utilise uniformed measurements of preventive care provision and elucidate the preventive care policy and procedural contexts in which services operate.

Review findings should be considered with caution due to the very low quality of cumulative evidence, suggesting future studies are likely to change the estimates; and in light of the following limitations. The majority of studies were categorised as high risk of bias. Data relied on self-report measures and medical record audit; where clinician self-report may over-estimate care provision, whilst audit data may reflect an underestimation of care provision; (Hrisos et al., 2009) suggesting client reported data may be a useful additional measure in future research where feasible (Hrisos et al., 2009). However, client report may also be susceptible to recall bias (Hrisos et al., 2009). Future research could seek to compare the accuracy of such measures in the context of preventive care in mental health services.

The variability in assessment and reporting of outcome measures by included studies limited comparability and resulted in many meta-analysis outcomes being based on a limited number of studies. Clinician reported data could not be pooled to create an estimate of care provision to all clients, rather the more conservative estimate of care provision to at least 50% of clients was analysed. Finally, heterogeneity remained substantial despite subgroup and sensitivity analyses suggesting between-study differences that could not be explored due to inconsistencies or insufficient reporting of study measures and characteristics.

Conclusion

This review found suboptimal provision of preventive care for tobacco smoking, harmful alcohol consumption, inadequate nutrition, and inadequate physical activity in mental health settings across all analysed care elements. These findings are important to mental health service clinicians and managers as they suggest current preventive care is suboptimal across risk behaviours and service types; suggesting a need to improve the quality of interventions to increase the provision of preventive care. Utilisation of consistent care provision and reporting measures across all mental health services, relating to care provided to individual clients, would facilitate further synthesis of the prevalence of care provision. Future studies need to investigate methods to increase clinician delivery of preventive care in mental health services, which have an organisational culture and service issues that pose idiosyncratic challenges to the field of implementation science (Sandstrom et al., 2015; Michie et al., 2007).

Systematic review protocol and registration

Systematic review of the prevalence of preventive care provision for chronic disease risk behaviours in mental health services. PROSPERO 2016:CRD42016049889 Available from: https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=49889