Disentangling the contribution of hospitals and municipalities for understanding patient level differences in one-year mortality risk after hip-fracture: A cross-classified multilevel analysis in Sweden.

BACKGROUND: One-year mortality after hip-fracture is a widely used outcome measure when comparing hospital care performance. However, traditional analyses do not explicitly consider the referral of patients to municipality care after just a few days of hospitalization. Furthermore, traditional analyses investigates hospital (or municipality) variation in patient outcomes in isolation rather than as a component of the underlying patient variation. We therefore aimed to extend the traditional approach to simultaneously estimate both case-mix adjusted hospital and municipality comparisons in order to disentangle the amount of the total patient variation in clinical outcomes that was attributable to the hospital and municipality level, respectively.
METHODS: We determined 1-year mortality risk in patients aged 65 or above with hip fractures registered in Sweden between 2011 and 2014. We performed cross-classified multilevel analysis with 54,999 patients nested within 54 hospitals and 290 municipalities. We adjusted for individual demographic, socioeconomic and clinical characteristics. To quantify the size of the hospital and municipality variation we calculated the variance partition coefficient (VPC) and the area under the receiver operator characteristic curve (AUC).
RESULTS: The overall 1-year mortality rate was 25.1%. The case-mix adjusted rates varied from 21.7% to 26.5% for the 54 hospitals, and from 18.9% to 29.5% for the 290 municipalities. The VPC was just 0.2% for the hospital and just 0.1% for the municipality level. Patient sociodemographic and clinical characteristics were strong predictors of 1-year mortality (AUC = 0.716), but adding the hospital and municipality levels in the cross-classified model had a minor influence (AUC = 0.718).
CONCLUSIONS: Overall in Sweden, one-year mortality after hip-fracture is rather high. However, only a minor part of the patient variation is explained by the hospital and municipality levels. Therefore, a possible intervention should be nation-wide rather than directed to specific hospitals or municipalities.

Introduction

Increasing demands for healthcare fueled by aging populations, increasing prevalence of multimorbid patients and technological advances combined with restricted financial resources constitute a major challenge for health care systems globally [1]. To address this challenging task many countries have started to streamline their healthcare systems by improving care and reducing length of stay in hospitals, which includes earlier discharge to care in community settings or at home with professional support (i.e., homecare) [2, 3]. Hip fracture care is an example of this development where surgery is performed at the hospital while postoperative rehabilitation and further care of potential sequelae are done in the primary health care sector. Therefore, evaluating the quality of care after hip fracture needs to simultaneously consider both the hospital and the municipality settings. When evaluating quality of care after hip fracture, one-year mortality is frequently used as an outcome quality indicator in many healthcare systems [4]. From this perspective, identifying unwarranted variability between not only hospitals but also municipalities has considerable health policy relevance [5-7].

It is well-established that multilevel modelling is a suitable methodology when comparing different types of cluster (e.g., hospital, municipalities) for both statistical and conceptual reasons [8-18]. However, existing multilevel model studies evaluating healthcare performance after hip fracture have primarily focused on hospital level variation [12, 19–22]. Less work has been carried out analysing the potential variability among municipalities [23]. Cross-classified multilevel analysis allows us to decompose the total individual level variation into its hospital and its municipality level components [24, 25]. To our knowledge no previous studies have applied measures of components of variance and of discriminatory accuracy to examine individual variation in one-year mortality with respect to both hospitals and municipalities. Therefore, we use cross-classified multilevel analysis to evaluate one-year mortality in all patients suffering from hip fracture in Sweden.

Population and methods

Databases and study population

This historical follow-up study was based on prospectively collected data available from medical registries in Sweden (about 9.3 million inhabitants by 2010) with free access to medical care. At birth or upon immigration, all citizens in Sweden are assigned a unique registration number through which all contact with the healthcare system is recorded. This allows unambiguous record linkage between registries.

We obtained information on diagnoses, medication use and mortality from the Swedish Patient Register, the Swedish Prescribed Drug Register [26], and from the Cause of Death Register [27] administered by the National Board of Health and Welfare. We obtained demographic and socioeconomic information from the Population Register [28] and the Longitudinal Integration Database for Health Insurance and Labour Market Studies (LISA) register (http://www.scb.se/lisa), which are administrated by Statistics Sweden. To ensure the anonymity of the subjects, the Swedish authorities transformed the personal identification numbers of the individuals into arbitrary numbers before delivering the research databases to us, and we linked the databases using the anonymized identification number.

We defined hip fracture as the presence of any discharge diagnosis with a fracture of the femur coded according the International Classification of Diseases, 10th revision (ICD codes) as fracture of the femoral neck (S72.0), pertrochanteric (S72.1) or subtrochanteric (S72.2). From the Swedish Patient Registry, we identified all 56,161 patients being 65 years or older and residing in Sweden by 31th December 2010 who were discharged from the Swedish hospitals between 2011 and 2014 with a diagnosis of hip (i.e., femoral) fracture and surgery code. Next, we excluded 1,162 patients because they were residing less than five years in Sweden (n = 173), they have missing information on education (n = 976) or erroneous information on death date (n = 2). We also excluded 11 patients treated at facilities with less than 10 hip fracture patients during the study period (Fig 1). Patients which previous hip fracture compose of 11.95% (6,575/48,424) of our population.

Fig 1

Flowchart patient inclusion.

Ethics and data accessibility

The Regional Ethics Review Board in southern Sweden (Dnr: 2014/856) as well as the data safety committees from the National Board of Health and Welfare and from Statistics Sweden approved the construction of the database used in this study.

The database we analysed is not publicly available for ethical and data safety reasons. However, the same dataset can be constructed by request to the Swedish National Board of Health and Welfare and Statistics Sweden after approval of the research project by an Ethical Committee and by the data safety committees of the Swedish Authorities. The study also needs to be performed in collaboration with Swedish researchers. [29]

Assessment of variables

Mortality

For each patient we ascertained all-cause mortality within one year from the admission date to hospital.

Sociodemographic and clinical characteristics of the patients

Part of the initial differences between hospitals or municipalities may relate to differences in case mix. To make the observational measurement of hospital and municipality effects as valid as possible we therefore adjusted for potential differences in patient sociodemographic and clinical characteristics as well as use of medication (Table 1).

Table 1

Characteristic of the hip fracture population.

Values are percentages (number of patients) if not otherwise indicated.

Overall 1-year mortality	25.08
Number of patients in the population	54,999
Number of hospitals	54
Number of municipalities	290
Median number of patients at the hospital (min–max)	1015.50 (158–3,724)
Median number of patients at the municipality (min–max)	189.65 (15–4,687)
Age group (years)
• 65–74	18.38 (10,108)
• 75–84	42.29 (23,258)
• > 85	39.33 (21,633)
Gender
• Men	30.87 (16,976)
• Women	69.13 (38,023)
Biomedical risk score for all-cause mortality
• Low	38.42 (21,132)
• Medium	11.94 (6,566)
• High	24.81 (13,645)
• Very high	24.83 (13,656)
Education
• Low education	85.31 (46,922)
• High education	14.69 (8,077)
Income
• Low	35.87 (19,728)
• Medium	41.33 (22,731)
• High	22.80 (12,540)
Migration status
• Immigrant	7.94 (4,365)
• Native (reference)	92.06 (50,634)
Cohabiting status
• Living alone (reference)	64.76 (35,620)
• Living together	35.24 (19,379)
Medication
• Bisphosphonates	0.63 (346)
• Analgesics	27.07 (14,886)
• Psycholeptics	59.90 (32,942)
• Psychoanaleptics	43.53 (23,939)

Sociodemographic characteristics

We use the age of the individuals in years as a continuous variable and included it in the analysis as a quadratic function. We categorized the patients as immigrants if they were born in another country and as natives if they were born in Sweden.

We calculated household individualized disposable income by dividing the total disposable income of the family of the patient by the number of family members, considering the different consumption weights of adults and children, according to Statistics Sweden (http://www.scb.se/lisa). We did so for the complete Swedish population in three occasions 2010, 2005 and 2000, and summarized the three occasion by computing the cumulative income. Finally, using the tertile values of the cumulative income distribution we divided the study population into high, medium and low income. In this form, we categorized each patient according his/her income category in the total population rather than the population of patients.

We classified education achievement as compulsory schooling of nine years or less, secondary education three years or less, and higher education. We defined low education achievement as compulsory school or less and use the value of the member of the family with the highest educational achievement for all member in the same family.

For defining those patients living alone, we first identified all patients who were cohabiting. That is, married couple, in a registered partnership or in an unregistered partnership with a common child as well as those living in a household with at least another adult. We group all other patients into the living alone category.

We combined sociodemographic characteristics into a single sociodemographic risk score for all-cause mortality. We did this via a logistic regression analysis modelling all-cause mortality as a function of age, gender, income, cohabiting, born in Sweden and education (see S1 Table). The individual predicted probabilities were then categorized into four groups by quartiles as low, medium, high, and very high. The low-risk score group was then used as the reference in the comparisons in the subsequent multilevel analyses.

Clinical characteristics

For the purpose of patient-mix adjustment and using previous knowledge on risk factors for mortality in patients with hip fractures [30, 31] as well as considering the variables included in the Charlson comorbidity index (CCI), we selected a number of diseases (ICD-10 codes) identified during the five years before hospital admission (see S2 Table). We then applied logistic regression to obtain a clinical risk score (i.e., individual predicted probability) of one-year mortality based on the following variables: Chronic kidney disease (N18), Acute myocardial infarction (I21), Heart failure (I50), Other peripheral vascular diseases (I73), Cerebrovascular diseases (I60-I69), Atherosclerosis- Aortic aneurysm and dissection (I70-I71), Dementia (F01-F03), Chronic lower respiratory diseases (J40-J47), Other disorders of the skin and subcutaneous tissue (L80-L99), Peptic ulcer (K27), Diseases of liver (K70-K77), Diabetes mellitus (E08-E13), Hemiplegia (G81), Neoplasms (C00-D49), Human Immunodeficiency Virus (HIV) (B20), osteoporosis (M80-M81), previous Hip fracture (S70-S72). Finally, we created four categories of clinical risk scores using the quartile values of the risk score distribution and considering the group with the lowest risk score as reference in the comparisons.

Use of medication

Using the Anatomical Therapeutic Chemical (ATC) classification from the Swedish Prescribed Drug Register we also obtained information on previous use of Analgesics (ATC cod: N02), Psycholeptics (N05), Psychoanaleptics (N06) and Bisphosphonates (M05BA) as these medications may influence the vital prognosis of the patients [32, 33].

Statistical analysis

We applied single-level (conventaional) and multilevel logistic regression analyses of discriminatory accuracy as decribed elsewhere [18]. We developed three consecutive logistic regression analyses to model one-year mortality.

The first model (model 1) is a single-level logistic regression including the socioeconomic risk score in four groups. This model aimed to evaluate the influence of patients’ demographical characteristic on one-year mortality.

The second (model 2) is also a single-level logistic regression that expands model 1 to include the risk score for clinical factors as well as use of medication.

In the final, third model (model 3), we expanded model 2 by adding two random effects, one for the hospital level and the other for municipality level. In this way the model was converted into a two-way cross-classified multilevel model with the patients nested within the 54 hospitals and the 290 municipalities. As described elsewhere [18], we calculated the hospital and municipality general contextual effects as expressed by the variance partition coefficient (VPC) which informs on the share of the total individual variance in the propensity of one-year mortality that is at the hospital (VPCH) and at the municipality level (VPCM).

For all models, we calculated the Area Under the receiving operator characteristics Curve (AUC) or C-statistics [34] as a measure of discriminatory accuracy. Hosmer and Lemeshow [35] suggest that an AUC of 0.70 to 0.80 could be considered as 'acceptable', 0.80 to 0.90 as 'excellent' and 0.90 or above as 'outstanding', while an area under the ROC curve of 0.50 suggests no discrimination between the outcome groups (i.e., similar as tossing a coin to decide group membership). In model 3 we separately calculated the AUC when adding the random effect for the hospital (AUC) and the random effect for the municipality (AUC). Then, we calculated the increment in the AUC [18] when going from model 1 to model 2 (AUCΔ2−1) and from model 2 to model 3 (AUCΔ3−2). The increment in AUC measures the improvement in the ability of the model to correctly classify individuals with or without the outcome (i.e., one-year mortality) when considering the hospital or the municipality of the patients. The AUC provides analogous information as the VPC [24] as both measures inform on the general contextual effect of the hospital and of the municipality levels in relation to the patients’ one-year mortality. As it has been explained elsewhere [12, 13, 18, 36–41] the general contextual effect, expresses the relevance of the hospital/municipality context for understanding patients’ differences in one-year mortality.

In model 3 we use the predicted hospital and municipality random effects (shrunken residuals) to obtain case-mix adjusted and reliability-weighted average mortality rates and their 95% credible intervals and created league tables illustrating the ranking of the hospitals and of the municipalities.

In the supplementary material S1 File we provided an extended explanation of the methodology.

Results

Characteristics of the hip fracture population

The Swedish cohort of hip fracture patients consists of 54,999 patients. The unadjusted population one-year mortality was 25.1%. The hip fracture patients were mainly women, and most of the patients were living alone. Table 1 describes the additional characteristics of the population.

Patient effects

High sociodemographic scores and high clinical scores were strong predictors for one-year mortality (Table 2). The AUC1 in model 1, which informs on the discriminatory accuracy of the sociodemographic information, had a value of 0.667 (95% CI: 0.662–0.672). Including the clinical score of the patients and use of medicines increased the AUC to 0.716 (95% CI: 0.711–0.720) (model 2). The use of analgesics, psycholeptics or psychoanaleptics was associated with an increased risk of one-year mortality (Table 2). However, use of bisphosphonates was not associated with one-year mortality.

Table 2

Analysis of 1-year mortality after hip fracture in the Swedish hospitals.

	Simple logistic regression analysis				Cross classified multilevel logistic regression analysis
	Model 1		Model 2		Model 3
Specific individual average effects
Sociodemographic RS
•Low	1.00		1.00		1.00
•Medium	1.88	(1.75–2.01)	1.74	(1.62–1.86)	1.71	(1.59–1.82)
•High	2.99	(2.80–3.19)	2.72	(2.55–2.91)	2.68	(2.52–2.84)
•Very high	5.68	(5.33–6.04)	5.34	(5.01–5.69)	5.29	(4.98–5.57)
Clinical RS
•Low			1.00		1.00
•Medium			1.18	(1.10–1.27)	1.17	(1.08–1.25)
•High			1.54	(1.45–1.62)	1.53	(1.45–1.61)
•Very high			2.67	(2.53–2.81)	2.66	(2.52–2.80)
Medication
Bisphosphonates			0.91	(0.69–1.19)	0.91	(0.67–1.17)
Analgesics			1.14	(1.08–1.20)	1.13	(1.07–1.19)
Psycholeptics			1.26	(1.20–1.32)	1.26	(1.20–1.32)
Psychoanaleptics			1.32	(1.26–1.38)	1.32	(1.27–1.37)
General contextual effects
Hospital variance					0.007	(0.002–0.013)
Municipality variance					0.002	(0.001–0.005)
VPCH hospital					0.2
VPCM municipality					0.1
AUC	0.667	(0.662–0.672)	0.716	(0.711–0.720)	0.718	(0.713–0.722)
AUCΔ2−1 (increment model 2- model 1)	Reference		0.049
AUCΔ3−2 (increment model 3- model 2)			Reference		0.002

RS = Risk score for all-cause mortality, VPC = Variance Partition Coefficient, AUC = Area under the receiver operating characteristic curve

1) Model 1: Simple logistic regression model including the socioeconomic risk score for all-cause mortality

2) Model 2: Simple logistic regression model including the socioeconomic and biomedical risk scores for all-cause mortality

3) Model 3: Cross-classified multilevel logistic regression model including the socioeconomic and biomedical risk scores for all-cause mortality and the hospitals and municipalities as random effects.

Hospital effects

In the period from 2011 to 2014, 54 hospitals treating patients with hip fracture were included in our study. These hospitals treated between 158 and 3,724 patients. The hospital unadjusted one-year mortality expressed as percentage varied between 19.6% and 29.8%. The league table obtained from the multilevel analysis shows that after adjustment for patient case-mix and municipality effects the hospital mortality rates extended from 21.7% to 26.5% (Fig 2). The general contextual effect of the hospital on the patients' risk for one-year mortality was, however, low as only 0.2% of the adjusted individual variation in one-year mortality laid at the hospital level. Also, the increase in the discriminative accuracy when adding the hospital level to the model including only patient level variables was only 0.002 units (model 2).

Fig 2

League table ranking the 54 hospitals according to their adjusted absolute risk of 1-year mortality with 95% confidence intervals obtained from the cross-classified multilevel model.

Municipality effects

After admission the hip fracture patients were discharged to 290 different municipalities in Sweden. The number of hip fracture patients in each municipality varied from 15 to 4,687 hip fracture patients in the three years. The unadjusted mortality rate expressed as percentage varies between 7.1% and 37.9%. After adjustment for the difference in case-mix, and the hospital effects in the multilevel analysis the municipality differences in one-year mortality extended between 18.9% and 29.5% (Fig 3). The general contextual effect of the municipalities on the patients' risk for one-year mortality was also low as only 0.1% of the total individual variance in the propensity of dying was at the municipality level. Also, the increase in the discriminative accuracy when adding the municipality level to the model including only patient level variables was only 0.002 units (model 2).

Fig 3

League table ranking the 290 municipalities according to their adjusted absolute risk of 1-year mortality with 95% confidence intervals obtained from the cross-classified multilevel model.

Discussion

In this nationwide population-based study of Swedish hip fracture patients, the crude one-year mortality rate was 25.1% during the period 2011–2014. Patient sociodemographic and clinical characteristics were predictors of one-year mortality (AUC = 0.718), whereas the cross-classified multilevel analysis revealed that the AUC of the regressions model including hospitals and municipality as random effects was only marginally higher compared to the single-level model. Accordingly, hospital and municipality level variation corresponded to less than 1% of the overall individual variation in the underlying propensity of death within one year.

For evaluating the performance of the hospital and municipality health care performance we considered two measures, the overall unadjusted one-year mortality rate after hip fracture and the size of the general contextual effects. In our study the overall unadjusted one-year mortality rate of 25.08% is comparable to a previous Danish register study, which in the period 2011 to 2014 reported a one-year mortality of 26% [42]. However, our mortality rate was higher than that found in a previous study on seven European countries, which found a mortality rate of 22.3% in Sweden in 2007 [43]. In this European study the lowest mortality, 19%, was observed in Italy. Those differences might be explained by differences in study design and changing comorbidity patterns across time as well as country differences in healthcare provision [42, 44].

After adjustment for patient case-mix, the size of the general context effect is assumed to inform on the influence of the hospital and municipality health care levels on the vital prognosis of the patients [24]. The VPCs we found in this study were small, being 0.1% for municipalities and 0.2% for hospitals. Meaning that once we have adjusted for patient characteristics, patients treated at the same hospital or at the same municipality have very little in common according to their propensity of death within one year. That is, there were very small differences between hospitals and between municipalities (se elsewhere for an extended explanation on this concept [24])

To our knowledge no previous studies have applied cross-classified multilevel analyses to hip fracture data and so a direct comparison is therefore not possible. However, our results are in line with a previous Danish multilevel study examining hospital variation in 30-day mortality among hip fracture patient (but ignoring municipality variation), which found a VPC for hospitals of 0.9%.

The AUC of the cross-classified multilevel analysis was slightly higher compared to the single-level model, which still indicated that the cross-classified multilevel analysis was a better model than the single-level model even though the VPC was low for both the hospital and the municipality levels. However, the AUCs of the models were overall moderate, ranging from 0.667 to 0.718.

In our study bisphosphonates were not associated with reduced mortality. In randomized controlled trials, use of bisphosphonates have been shown to be effective for preventing both osteoporotic fractures [45-47] and mortality after hip fracture [48]. The protective effect of bisphosphonates on mortality after hip fracture has also been observed in numerous observational studies. However, recent research has suggested that the reduction in mortality occurs within the first weeks after treatment and, therefore, could express confounding [33].

Our study was strengthened by the cross-classified multilevel study design, as well as by the nationwide population-based coverage with complete follow-up on one-year mortality due to linkage to Swedish registries. However, our study only analysed the Swedish hospitals and municipalities and, therefore, cannot be generalized to other country contexts.

The treatment, care and rehabilitation of hip fracture patients at both hospitals and municipalities are fundamentally important. We found that hospital and mortality differences in average mortality risks were only a minor share of the total individual variance in the propensity to die or that, in other words, adding the hospital and municipality level does not increase the discriminatory accuracy obtained by patient level information only. This means that the hospital and municipality performances are homogenous overall in the country. Therefore, our results indicate that we will not lower mortality after hip fracture in Sweden by focusing on specific hospitals or specific municipalities with high average mortality. Rather, special efforts to reduce mortality after hip fracture should be focused on vulnerable patient groups of hip fracture patients wherever they are. From this reason, however, hospitals and municipalities with a higher number of vulnerable patients than other will need more intense interventions. Future multilevel studies among hip fracture patients need to investigate differences in provided care and intermediate outcomes such as complications [49].

In conclusion, using cross-classified multilevel regression analysis, we observed that overall in Sweden, one-year mortality after hip-fracture at 25.1% was rather high during the period 2011–2014. We also observed that the average one-year adjusted mortality varied between 21.7% and 26.5% for hospitals and between 18.9% and 29.5% for municipalities. However, while the patient socioeconomic and clinical characteristics appears relevant for predicting mortality, only a minor part of the patient variation was explained by the hospital and municipality levels. Therefore, future interventions should focus on identifying high risk patient groups and be nationwide rather than directed at specific hospital or municipalities.

Sociodemographic risk score.

(DOCX)

Click here for additional data file.

Clinical risk score.

(DOCX)

Click here for additional data file.

Statistical appendix.

(DOCX)

Click here for additional data file.

7 Apr 2020

PONE-D-20-07054

Disentangling the contribution of hospitals and municipalities for understanding patient level differences in one-year mortality risk after hip-fracture: a cross-classified multilevel analysis in Sweden

PLOS ONE

Dear Dr. Kristensen,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The study has been reviewed by one expert and the editor, who consider the study of merit and well presented. There are some specific suggestions to be considered by the authors, including the possibility of risk clustering and how this could affect the analysis. Further, additional descriptive parameters of the population could be of interest for the analysis.

We would appreciate receiving your revised manuscript by May 22 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Pablo Garcia de Frutos

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3. Thank you for stating the following in the Acknowledgments Section of your manuscript:

"we thank the Health Research Fund of Central Denmark Region Denmark and the Swedish Research Council for supporting this work https://www.swecris.se/betasearch/details/project/201701321VR and Vetenskapsrådet projekt-id: 2017-01321."

We note that you have provided funding information that is not currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

"Juan Merlo

Vetenskapsrådet projekt-id: 2017-01321

https://www.swecris.se/betasearch/details/project/201701321VR

NO

Pia Kjær Kristensen

The Health Research Fund of Central Denmark Region

A869

https://www.rm.dk/sundhed/faginfo/forskning/region-midtjyllands-sundhedsvidenskabelige-forskningsfond/

NO"

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This paper reports an analysis of national data on hip fracture mortality in older adults using a cross-classified multilevel framework to examine potential cluster effects by hospital and by municipality. Strengths of this study include the use of a large and comprehensive dataset and a sophisticated analytical approach. Overall, this is a strong contribution and my comments are relatively minor.

1. Additional supplementary information regarding the calculation of the sociodemographic and clinical risk scores would be valuable -- e.g., results for the logistic regression models from which these scores were derived would be good to include in the supplemental materials, including model fit and the parameter estimates used to compute the individual risk scores.

2. Because the cluster VPC statistics were calculated only after controlling for patient sociodemographic and clinical factors, there is some risk of under-estimating geographic disparities due to clustering of sociodemographic risk factors and patient comorbidity rates. It is likely that these risk factors are themselves clustered by municipality (and potentially to a lesser extent by hospital) based on local deprivation. The analytic reasons for taking the approach you used is explained clearly, but the issue of risk clustering should be briefly addressed in the discussion section. In particular, the suggestion that interventions aimed at improving mortality outcomes should focus on individual factors at a national level rather than on focusing on particular low-performing hospitals and municipalities should be somewhat tempered by the fact that certain specific places may require more focussed intervention because they include clusters of high-risk patients.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

30 Apr 2020

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Reply: We thank for pointing out this example of author guidelines. On the title page we have added the ¶ symbol for 1st set of equal contributors. We have corrected the headings to be written in sentence case instead of capitalize. In addition, we have deleted the graphic images and space in the tables.

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

Reply: We apologize for the lacking information on data approval and access. The dataset are only available from the Swedish National Board of Health and Welfare and Statistics Sweden upon request following an ethical approval. The database was constructed after approval from the Ethical Committee in Sweden (https://etikprovningsmyndigheten.se/) and from the data safety committees of the Swedish National Board of Health and Welfare (https://www.socialstyrelsen.se/) and Statistics Sweden (https://www.scb.se/en/).

3. Thank you for stating the following in the Acknowledgments Section of your manuscript:

"we thank the Health Research Fund of Central Denmark Region Denmark and the Swedish Research Council for supporting this work https://www.swecris.se/betasearch/details/project/201701321VR and Vetenskapsrådet projekt-id: 2017-01321." We note that you have provided funding information that is not currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form. Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

"Juan Merlo

Vetenskapsrådet projekt-id: 2017-01321

https://www.swecris.se/betasearch/details/project/201701321VR

NO

Pia Kjær Kristensen

The Health Research Fund of Central Denmark Region

A869

https://www.rm.dk/sundhed/faginfo/forskning/region-midtjyllands-sundhedsvidenskabelige-forskningsfond/

NO"

Reply: We thank the editor for pointing this out. We have deleted the phrases regarding funding statement in the acknowledgement section in the manuscript page 19 from line 340 to line 343. The original expression in the acknowledgement section (“the Swedish Research Council for supporting this work https://www.swecris.se/betasearch/details/project/201701321VR and Vetenskapsrådet projekt-id: 2017-01321”) was unfortunately confusing as it may represent two different funding statements. However, the “Swedish Research Council” is the English name of Swedish “Vetenskapsrådet”. So the funding statement for Juan Merlo is:

Swedish Research Council (i.e., Vetenskapsrådet) project-id: 2017-01321. https://www.swecris.se/betasearch/details/project/201701321VR

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Reviewer #1: This paper reports an analysis of national data on hip fracture mortality in older adults using a cross-classified multilevel framework to examine potential cluster effects by hospital and by municipality. Strengths of this study include the use of a large and comprehensive dataset and a sophisticated analytical approach. Overall, this is a strong contribution and my comments are relatively minor.

Reply: We thank the reviewer for the kind remarks.

1. Additional supplementary information regarding the calculation of the sociodemographic and clinical risk scores would be valuable -- e.g., results for the logistic regression models from which these scores were derived would be good to include in the supplemental materials, including model fit and the parameter estimates used to compute the individual risk scores.

Reply: We agree with the reviewer that calculation of the sociodemographic and clinical risk scores would be valuable for the interested reader. We have, therefore, as suggested by the reviewer, updated the Statistical appendix (labeled S1-S2 Table) to include information on the beta coefficients for the included variables in the Socioeconomic and clinic risk score equations as well as proving AUC values.

2. Because the cluster VPC statistics were calculated only after controlling for patient sociodemographic and clinical factors, there is some risk of under-estimating geographic disparities due to clustering of sociodemographic risk factors and patient comorbidity rates. It is likely that these risk factors are themselves clustered by municipality (and potentially to a lesser extent by hospital) based on local deprivation. The analytic reasons for taking the approach you used is explained clearly, but the issue of risk clustering should be briefly addressed in the discussion section. In particular, the suggestion that interventions aimed at improving mortality outcomes should focus on individual factors at a national level rather than on focusing on particular low-performing hospitals and municipalities should be somewhat tempered by the fact that certain specific places may require more focussed intervention because they include clusters of high-risk patients.

Reply: We thank the reviewer for bringing our attention to the possible risk of under-estimating geographic disparities due to clustering of sociodemographic risk factors and patient comorbidity rates. We have now estimated the VPC statistics for the empty model for considering the maximal clustering. The empty model had a VPC values for both hospitals and municipalities of 0.007. We therefore, think that the risk of under-estimating geographic disparities is minor. Besides, even if the empty VPC were much higher and reduced when considering the composition of patients (because some hospitals include clusters of high-risk patients), the directing resources to high-risk patients on a national basis would automatically allocate more resources to those hospitals and municipalities with more high-risk patients. We have therefore, elaborated the paragraph on page 17 from line 312 to line 318 in the Discussion section to the flowing:

We found that hospital and mortality differences in average mortality risks were only a minor share of the total individual variance in the propensity to die or that, in other words, adding the hospital and municipality level does not increase the discriminatory accuracy obtained by patient level information only. This means that the hospital and municipality performances are homogenous overall in the country. Therefore, our results indicate that we will not lower mortality after hip fracture in Sweden by focusing on specific hospitals or specific municipalities with high average mortality. Rather, special efforts to reduce mortality after hip fracture should be focused on vulnerable patient groups of hip fracture patients wherever they are. From this reason, however, hospitals and municipalities with a higher number of vulnerable patients than other will need more intense interventions.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

19 May 2020

Disentangling the contribution of hospitals and municipalities for understanding patient level differences in one-year mortality risk after hip-fracture: a cross-classified multilevel analysis in Sweden

PONE-D-20-07054R1

Dear Dr. Kristensen,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Pablo Garcia de Frutos

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

22 May 2020

PONE-D-20-07054R1

Disentangling the contribution of hospitals and municipalities for understanding patient level differences in one-year mortality risk after hip-fracture: a cross-classified multilevel analysis in Sweden

Dear Dr. Kristensen:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Pablo Garcia de Frutos

Academic Editor

PLOS ONE