The impact of the resumption of in-person school attendance on COVID-affected child abuse and neglect trends in Florida.

BACKGROUND: The COVID-19 pandemic has led to fewer child abuse and neglect (CAN) hotline calls, CAN investigations, and foster care entries across the U.S.
OBJECTIVE: To determine if there were decreases in CAN hotline calls, CAN investigations, foster care entries, and foster care exits in Florida (the largest among the few states that publishes monthly public data on the four areas) after the pandemic began, and to determine if there was any amelioration of these trends in Florida once schools had reopened for in-person learning.
METHODS: Secondary data analyses of administrative child welfare data from January 2010 to June 2021 from Florida were used. Spline regression equations were calculated for CAN hotline calls, CAN investigations, foster care entries, and foster care exits during three periods: 1) before the COVID-19 pandemic (January 2010 to February 2020); 2) after the pandemic, but before schools reopened (March 2020 to October 2020); and 3) after the pandemic and after schools reopened (November 2020 to June 2021).
RESULTS: Prior to the pandemic, there was a non-statistically significant increase of 1.35 hotline calls per 100,000 children per month (p = .478), a statistically significant increase of 0.01 investigations per 100,000 children per month (p < .001), a non-statistically significant increase of 0.01 foster care entries per 100,000 children per month (p = .415), and a statistically significant increase of 0.03 foster care exits per 100,000 children per month (p < .05). Once the pandemic started in March 2020, there were statistically significant decreases of 136.02 hotline calls per 100,000 children per month (p < .001), 102.84 investigations per 100,000 children per month (p < .001), 6.32 foster care entries per 100,000 children per month (p < .001) and 5.75 foster care exits per 100,000 children per month (p < .01). Once all schools reopened for in-person learning in November 2020, there continued to be statistically significant decreases of 47.86 hotline calls per 100,000 children per month (p < .05), 6.38 foster care entries per 100,000 children per month (p < .001) and 6.53 foster care exits per 100,000 children per month (p < .001). This suggests that there were an estimated 34,374 fewer CAN hotline calls, 2338 children who did not enter foster care, and 2587 youth residing in foster care (YRFC) whose foster care exits were delayed. The delay in foster care exits suggests that YRFC in Florida had stayed a combined cumulative equivalent of 477.1 years longer in care.
CONCLUSION: With the COVID-19 variants like Omicron continuing to wreak havoc in Florida, there will be a continued trend of decreasing CAN hotline calls, foster care entries and foster care exits even with in-person learning in all Florida schools. A real-time interoperable data system utilizing real-time predictive analytics must be developed in concert with the development of leaders and executives with advanced degrees in child welfare organizations who are able to maximize information from such systems.

Introduction

Child abuse and neglect (CAN) is a serious public health issue leading to CAN hotline calls on 37.4 million children nationally from Federal Fiscal Years (FFY) 2015 to 2019 inclusive (U.S. Children’s Bureau, 2021). In the same time period, there were CAN investigations on 17.3 million children, of whom approximately 14.0% were provided preventative services and another 16.8% were substantiated for CAN (U.S. Children’s Bureau, 2021). The largest sources of CAN reports came from school personnel (21.0%), law enforcement (19.1%) and medical professionals (11.0%).

COVID-19 was first detected in the United States in January 2020 (Auger et al., 2020). While national emergencies have led to increased child abuse nationally (Seddighi et al., 2021), COVID-19 has led to decreased child abuse reports (Rapoport et al., 2021), missed child abuse investigations across the U.S. (Nguyen, 2021a), and a decrease in the number of children who entered foster care in some states (Musser et al., 2021). Research has suggested that decreases in CAN reports were partly due to significant decreased contacts between children and school officials (Rapoport et al., 2021) due to the closure of schools (Auger et al., 2020) as schools are the largest source of CAN reports (Humphreys et al., 2020).

Florida had the ninth largest number of CAN hotline calls and sixth largest number of CAN victims in the U.S. in FFY 2019 (U.S. Children’s Bureau, 2021). In Calendar Year 2019, there were over 300,000 child abuse hotline calls in Florida, which led to 34,837 children who were confirmed to have been abused or neglected (Florida Department of Children and Families, 2022). The first COVID cases were detected in Florida in March 2020 and all school districts in the State suspended in-person learning by March 20, 2020 (Doyle et al., 2021). Florida was also one of the first states to reopen most of its school districts to in-person learning in August 2020 and by November 2020, all school districts had reopened to in-person learning.

Hypotheses

Given that the state of Florida has among the highest numbers of substantiated cases of child abuse and neglect in the United States, and that it was one of the first states to resume in-person school during the COVID-19 pandemic, the current study examined the impact of COVID-19 pandemic in Florida up to and including June 2021. The hypotheses are that: 1) when the pandemic began, there would be significant decreases in CAN hotline calls, CAN investigations, foster care entries, and foster care exits in Florida; and 2) once in-person attendance resumed in all school districts in Florida in November 2020, the decreases would subside in the number of CAN hotline calls, CAN investigations, foster care entries, and foster care exits.

Methods

The approximate monthly number of CAN hotline calls, children screened-in for CAN investigations, foster care entries, and foster care exits were identified from secondary administrative child welfare data from January 2010 to June 2021 from Florida for those youth under 18 years of age (Florida Department of Children and Families, 2022). Foster care in Florida can include placements with relatives (kinship care), in licensed foster homes, group homes, and juvenile detention facilities (Florida Department of Children and Families, 2022).

Regression analyses were utilized to see if there were any statistically significant changes to each of these four categories. Spline regression modeling has to be utilized as a traditional linear regression cannot be used to determine trends when there were two major events, the beginning of the pandemic and schools reopening to in-person learning, that could significantly change the slope of the line after each of these events suggesting a non-linearity (Marsh & Cormier, 2001). In order to take two significant events into account, a modified linear regression equation was generated:

Based on the above, “y” represents the final value of the dependent variable, “a” represents the y-intercept, “b ” represents the slope before the pandemic began, “x” represents the value of the independent variable, “b ” represents the slope between the beginning of the pandemic and before schools reopened to in-person learning, “x k1 ” value at the start of the pandemic (i.e., 123 or the 123rd month or March 2020), “x ” represents the dummy variable created if it was after the pandemic, but before schools reopened, “b ” represents the slope after all schools reopened for in-person learning, “x ” value at the school reopening (i.e., 131 or the 131st month or November 2020) when all children in Florida had returned to in-person learning (Doyle et al., 2021) and “x ” represents the dummy variable created if it was after schools reopened. In essence, a regression equation was calculated for the first slope representing the line before the pandemic. A second regression equation was calculated representing the second slope of the line once the pandemic began, but before schools started. A statistical analysis was then done to see if there was a statistically significant difference between the second slope compared to the first slope. Then a third regression equation was calculated (representing the third slope of the line once in-person learning resume). A statistical analysis was also done to see if there was a statistically significant difference between the third slope compared to the second slope.

Linear regression models assume that the residuals of the variables are independent of each other (Brower & Jeong, 2007). Autocorrelation must be addressed when examining time-series data used to generate traditional linear regression models (Graddy & Wang, 2007) and has been successfully used to examine child welfare trends during the pandemic (Nguyen, 2021a). A Durbin–Watson statistic (a measure of autocorrelation) was conducted which could range from 0 to 4 (Graddy & Wang, 2007). A Durbin–Watson statistic closer to 0 represents positive autocorrelation, a statistic of 2 represents no autocorrelation while a statistic close to 4 suggests negative autocorrelation. If there is autocorrelation, the Prais–Winsten method using an autoregressive two-stage model or AR(1) was conducted. The Prais–Winsten method estimates a weighted least-squares regression, then estimates the autocorrelation parameter for an AR(1) process in the error terms (Bitle & Zavodni, 2002). The AR(1) reduces the autocorrelation. SPSS was utilized to calculate the regression models accounting for autocorrelation (including both the Durbin–Watson statistic and the Prais–Winsten method).

Once numbers were calculated, they were transformed into the number of cases per 100,000 children using the estimate of the current number of children in Florida, or 3,704,499 (U.S. Census Bureau, 2022). In essence, for each of the resultant number of CAN hotline calls, children screened-in for CAN investigations, foster care entries, and foster care exits, the number was divided by 3,704,499 and multiplied by 100,000.

The estimated number of youth residing in foster care (YRFC) who stayed longer than predicted was then calculated. The estimated average in care for each month (Florida Department of Children and Families, 2022) was then multiplied by the number of children for that month, and total estimates of additional foster care length of time were then calculated per month.

Results

General outcomes

The monthly changes in Florida can be seen in Table 1 for CAN hotline calls, CAN investigations, foster care entries, and foster care exits. Prior to the pandemic, the average monthly hotline calls ranged from 19,800 (June) to 24,816 (May). After the pandemic, the month with the lowest CAN hotline calls was April 2020 (17,176). Once all schools returned to in-person learning, the month with the lowest CAN hotline calls was June 2021 (21,689).

Table 1

From January 2010 to June 2021, the number of child abuse and neglect hotline calls, investigations, foster care entries, and foster care exits in Florida.

	CY2010	CY2011	CY2012	CY2013	CY2014	CY2015	CY2016	CY2017	CY2018	CY2019	CY2020	CY 2021	Average Number Prior to Pandemica
Hotline Calls
January	15,102	21,383	21,412	22,854	22,331	16,461	22,211	24,097	25,485	26,000	26,235	24,022	22,143
February	16,457	19,426	21,846	21,800	21,569	15,784	22,855	23,975	26,062	26,413	26,121	24,961	22,028
March	21,919	21,412	22,539	21,477	22,708	17,155	24,356	25,705	27,214	25,232	22,594	26,741	22,972
April	21,160	21,846	22,307	24,944	24,837	18,686	24,719	25,762	28,748	29,073	17,176	27,236	24,208
May	21,522	22,539	23,541	25,844	25,264	19,107	26,064	27,555	28,432	28,292	19,115	26,332	24,816
June	18,297	22,307	18,091	19,513	20,436	15,572	21,497	20,992	20,915	20,378	19,897	21,689	19,800
July	18,117	23,541	18,827	19,723	20,404	15,720	20,325	20,748	21,626	20,706	20,423		19,974
August	18,875	18,091	20,690	21,660	21,344	16,305	24,165	24,883	26,066	24,170	21,646		21,625
September	20,844	18,827	21,359	23,130	23,585	18,013	25,064	21,452	26,859	25,554	24,405		22,469
October	21,906	20,690	24,546	24,710	24,044	18,415	23,814	27,512	27,858	28,152	26,696		24,165
November	19,711	21,359	20,497	21,194	20,013	15,415	23,309	25,465	24,189	23,503	22,650		21,466
December	17,424	24,546	19,545	20,499	20,433	15,731	22,164	23,934	22,804	22,928	22,385		21,001
Investigations
January	14,911	17,691	16,458	16,230	16,461	17,853	17,440	18,728	18,048	17,828	16,510	15,264	17,105
February	15,660	17,476	16,942	15,959	15,784	16,934	17,699	18,420	18,270	17,960	16,145	15,690	17,023
March	18,754	20,985	17,624	15,855	17,155	18,620	18,757	19,534	18,946	17,175	14,305	16,880	18,341
April	18,181	19,929	17,403	18,396	18,686	19,412	19,016	19,228	20,256	19,749	10,941	17,107	19,026
May	18,642	20,496	18,266	18,794	19,107	18,933	20,319	20,776	19,879	19,067	12,537	16,666	19,428
June	15,834	16,765	14,006	14,202	15,572	15,931	16,480	15,807	14,610	13,898	12,871	13,901	15,311
July	15,607	15,468	14,450	14,074	15,720	15,805	15,572	15,448	15,140	13,912	13,270		15,120
August	16,283	17,585	15,570	15,707	16,305	16,371	18,518	18,449	18,205	15,991	13,797		16,898
September	17,870	18,302	15,872	16,806	18,013	17,408	19,216	15,710	18,702	16,638	15,485		17,454
October	18,893	17,849	17,899	18,041	18,415	18,360	18,459	19,643	19,351	18,181	16,763		18,509
November	16,868	17,409	14,860	15,692	15,415	17,080	18,084	17,987	16,875	15,233	14,394		16,550
December	14,916	15,117	13,872	15,256	15,731	17,119	17,164	16,862	15,812	14,679	14,244		15,653
Foster Care Entries
January	989	1214	1253	1256	1272	1447	1449	1380	1362	1238	1164	979	1275
February	1108	1234	1364	1023	1138	1527	1383	1287	1316	1142	1154	1064	1243
March	1236	1577	1363	1167	1305	1589	1350	1497	1397	1257	1128	1193	1374
April	1285	1351	1362	1184	1559	1698	1448	1398	1270	1265	972	1100	1382
May	1297	1423	1400	1114	1524	1407	1523	1564	1381	1358	1098	1055	1399
June	1380	1278	1129	1028	1158	1338	1451	1325	1289	1162	1108	1092	1254
July	1195	1171	1236	1178	1245	1344	1308	1342	1281	1143	1135		1244
August	1200	1559	1270	1403	1367	1491	1647	1587	1536	1352	1040		1441
September	1199	1318	1151	1242	1240	1443	1581	1075	1204	1232	1052		1269
October	1264	1351	1237	1448	1629	1587	1475	1448	1290	1313	1125		1404
November	1079	1203	1048	1134	1162	1275	1371	1346	1149	947	910		1171
December	944	1231	996	1128	1291	1370	1261	1133	1053	941	1064		1135
Foster Care Exits
January	851	921	921	969	902	876	998	993	1140	1023	949	714	958
February	990	1006	1048	1085	835	939	1290	1111	1113	1048	1000	861	1042
March	1249	1081	1185	1182	1094	1223	1324	1280	1157	1100	991	1036	1188
April	1095	990	1171	1203	1039	1131	1252	1186	1176	1180	1097	1032	1142
May	878	1012	1250	1285	1098	1029	1249	1280	1242	1306	940	971	1163
June	1427	1403	1436	1491	1377	1428	1566	1582	1573	1283	1280	1236	1457
July	929	847	1079	1105	931	1169	1179	1128	1181	1127	882		1068
August	1217	1330	1424	1131	1256	1305	1350	1416	1427	1230	957		1309
September	1036	998	916	986	937	1085	1128	886	918	902	1012		979
October	956	1006	1127	1068	1199	1168	1178	1222	1191	1111	1020		1123
November	1335	1376	1356	1189	1228	1287	1421	1566	1440	1501	1061		1370
December	1213	1453	1475	1152	1257	1287	1342	1367	1261	1099	1090		1291

Average number for January and February includes information from CY 2010 to CY 2020, and for the other months includes information from CY 2010 to CY 2019.

The average monthly investigations prior to the pandemic ranged from 15,120 (July) to 19,428 (May). After the pandemic, the month with the lowest investigations was April 2020 (10,941) and once schools returned to in-person learning, the month with the lowest investigations was June 2021 (13,901).

The average monthly foster care entries prior to the pandemic ranged from 1135 (December) to 1441 (May). After the pandemic, the month with the lowest foster care entries was May 2020 (2129). After the pandemic, the month with the lowest foster care entries was November 2020 (910). Once schools began in-person, the month with the lowest foster care entries was January 2021 (979).

The average monthly foster care exits prior to the pandemic ranged from 958 (January) to 1457 (June). After the pandemic, the month with the lowest foster care entries was July 2020 (882). Once schools began in-person, the month with the lowest foster care exits was January 2021 (714).

Regression analysis

As seen in Table 2 , prior to the pandemic, there were non-statistically significant increases of 50.1 hotline calls per month (p = .478) in Florida, and a statistically significant (p < .001) increase of less than one investigation (0.5) per month. There was a non-statistically-significant increase of 0.5 foster care entries per month (p = .415), and a statistically significant increase of 1.0 foster care exits (p < .05) in Florida prior to the pandemic.

Table 2

Regression equations for child abuse and neglect (CAN) hotline calls, CAN investigations, foster care entries, and foster care exits in Florida: 1) before the Pandemic (January 2010 to February 2020); 2) once the Pandemic began, but before school starting (March 2020 to October 2020); and 3) after the Pandemic once schools reopened (November 2020 to June 2021).

Area	R2	df	Durbin-WatsonStatistic	Regression Equation Before the PandemicJan 2010 to Feb 2020	Regression Equation Once the Pandemic Began, but Before All Schools ReopenedIn-PersonMar 2020 to Oct 2020	Regression Equation After the Pandemic, and once Schools ReopenedNov 2020 to June 2021
Hotline Calls	0.312	3133	1.737	y = 19,684.903 + 50.135× − 5089.066(x - 123)*(0) -1823.105(x - 131)*(0)ory = 19,684.903 + 50.135×p = .478	y = 19,684.903 + 50.135× − 5089.066(x - 123)*(1) -1823.105(x - 131)*(0)ory = 645,640.021–5038.931 xp < .001	y = 19,684.903 + 50.135× − 5089.066(x - 123)*(0) -1823.105(x - 131)*(1)ory = 258,511.6581772.97×p < .05
Investigations	0.133	3133	1.788	y = 16,906.020 + 0.4870× − 3810.299(x - 123)*(0) -1994.954(x - 131)*(0)ory = 16,906.020 + 0.4870×p < .001	y = 16,906.020 + 0.4870× − 3810.299(x - 123)*(1) -1994.954(x - 131)*(0)ory = 485,572.797–3809.812 xp < .001	y = 16,906.020 + 0.4870× − 3810.299(x - 123)*(0) -1994.954(x - 131)*(1)ory = 262,285.362–1994.467×p = .083
Foster Care Entries	0.099	3133	2.137	y = 1267.008 + 0.467× − 234.628(x - 123)*(0) -254.578(x - 131)*(0)ory = y = 1267.008 + 0.467×p = .415	y = 1267.008 + 0.467× - 234.628(x - 123)*(1) -254.578(x - 131)*(0)ory = 30,126.252–234.161×p < .001	y = 1267.008 + 0.467× − 234.628(x - 123)*(0) -254.578(x - 131)*(1)ory = 34,616.726–254.111×p < .001
Foster Care Exits	0.127	3133	1.983	y = 1110.287 + 0.994× − 214.184(x - 123)*(0) -242.934(x - 131)*(0)ory = 1110.287 + 0.994×p < .05	y = 1110.287 + 0.994× − 214.184(x - 123)*(1) -242.934(x - 131)*(0)ory = 27,454.919–213.190×p < .01	y = 1110.287 + 0.994× − 214.184(x - 123)*(0) -242.934(x - 131)*(1)ory = 32,934.641–241.940×p < .001

After the pandemic began, but before schools fully opened in-person in November 2020, there were statistically significant decreases of 5038.9 hotline calls per month (p < .001) and statistically significant decreases of 3809.8 investigations per month (p < .001) in Florida. There was a statistically significant decrease of 234.2 foster care entries per month (p < .001) and a statistically significant decrease of 213.2 foster care exits per month (p < .01).

After all schools allowed in person attendance in Florida in November 2020, there were statistically significant decreases of 1772 hotline calls per month (p < .05), but a non-statistically significant decrease of 1994.5 investigations per month (p = .083). There was a statistically significant decrease of 254.1 foster care entries per month (p < .001) and a statistically significant decrease of 241.9 foster care exits per month (p < .001) once in-person school attendance was allowed in all school districts in Florida.

Significant rates

As seen in Table 3 , once the pandemic began, the rates of CAN hotline calls went from a non-significant increase of 1.35 calls per 100,000 children to a significant decrease of 136.02 hotline calls per 100,000 children, and to a significant decrease of 47.86 hotline calls per 100,000 children once school attendance began in person. The rate of CAN investigations went from a significant increase of 0.01 investigations per 100,000 children to a significant decrease of 102.84 investigations per 100,000 children once the pandemic began. Once the pandemic began, the rates of foster care entries went from a non-significant increase of 0.01 foster care entries per 100,000 children to a significant decrease of 6.32 foster care entries per 100,000 children, and to a significant decrease of 6.38 foster care entries per 100,000 children once school reopened for in-person learning. The rates of foster care exits went from a significant increase of 0.03 foster care exits per 100,000 children prior to the pandemic to a significant decrease of 5.75 foster care exits per 100,000 children once the pandemic began, and to a significant decrease of 6.53 foster care exits per 100,000 children once school attendance began in person.

Table 3

Summary of outcomes for child abuse and neglect (CAN) hotline calls, CAN investigations, foster care entries, and foster care exits as rates per 100,000 children in Florida: 1) before the Pandemic (January 2010 to February 2020); 2) once the Pandemic began, but before all school districts re-opened in person (March 2020 to October 2020); and 3) after the Pandemic once all school districts re-opened in person (November 2020 to June 2021).

Outcome	Outcomes per 100,000 Children per MonthBefore the PandemicJan 2013 to Feb 2020	Outcomes per 100,000 Children per MonthOnce the Pandemic Began, but BeforeSchools ReopenedIn-PersonMar 2020 to Oct 2020	Outcomes per 100,000 Children per MonthAfter the Pandemic,and onceSchools ReopenedIn-PersonNov 2020 to Jun 2021
CAN Hotline Calls	Increase of 1.35 Calls	Decrease of 136.02 Calls***	Decrease of 47.86 Calls*
CAN Investigations	Increase of 0.01 Investigations***	Decrease of 102.84 Investigations***	Decrease of 53.84 Investigations
Foster Care Entries	Increase of 0.01 Foster Care Entries	Decrease of 6.32 Foster Care Entries***	Decrease of 6.38 Foster Care Entries***
Foster Care Exits	Increase of 0.03 Foster Care Exits*	Decrease of 5.75 Foster Care Exits**	Decrease of 6.53 Foster Care Exits***

⁎p < .05, ⁎⁎p < .01, ⁎⁎⁎p < .001.

Missed hotline calls/investigations/foster care entries/foster care exits

As seen in Table 4 , there were approximately 34,374 fewer CAN hotline calls and 32,955 fewer investigations from March 2020 to June 2021 than the 12 months prior to the start of the pandemic. For the same period, there were an estimated additional 2338 children who would have had substantiated CAN and would have been removed into foster care, but were not. The overall lifetime economic impact of each new foster care case in 2022-adjusted dollars is estimated to be $271,043 (Fang et al., 2012). This would suggest a lifetime economic cost of $633.7 million of the estimated 2338 children who were missed for substantiated CAN and not removed from foster care.

Table 4

Difference in the number of children the current month and the 12 months prior to the start of the pandemic for CAN Hotline Calls, Foster Care Entries, and Foster Care Exits in Florida showing those that had decreased numbers

	Hotline Calls		Investigations		Foster Care Entries		Foster Care Exits
Month	2020	2021	2020	2021	2020	2021	2020	2021
January		2213		1246		185		235
February		1160		455		90		139
March	2638		2870	295	129	64	109	64
April	11,897		8808	2642	293	165	83	148
May	9177		6530	2401	260	303	366	335
June	481		1027		54	70	3	47
July	283		642		8		245
August	2524		2194		312		273
September	1149		1153		180
October	1456		1418		188		91
November	853		839		37		440
December	543		435				9
Cumulative Total		34,374		32,955		2338		2587

There were 2587 YRFC whose foster care exits were delayed (Table 4) who stayed a cumulative total of 5725.7 months in care (Table 5 ). This would suggest that children were staying an average of 2.2 months longer in foster care once the pandemic began. In essence, YRFC in Florida had stayed a combined cumulative equivalent of 477.1 years longer in care. As the average caseload is 10 (Florida Department of Children and Families, 2021), this would suggest the equivalent of 36.43 child protective services workers. The salary of a child protective services worker in Florida is about $39,600 (Florida Department of Children and Families, 2021). With an estimated 50% in benefits, this would mean that a worker makes about $59,400 in Florida. With the equivalent of 36.43 child protective services workers, a delay would translate to a cost of approximately $2,834,217 over the 28-month period (from March 2020 to June 2021) or $101,222 per month.

Table 5

Estimated number of excess youth-months in foster care in Florida from March 2020 to June 2021.

Month	Referent Month	Referent Month Average Time in Care(a)	Exit Month Average Time in Care(b)	Difference(b-a)(c)	Number of Youth Who Stayed Longer than Average(d)	Excess Youth-Months in Care(c*d)
Mar-20	Mar-19	17.11	18.35	1.24	109	135.16
Apr-20	Apr-19	17.20	19.20	2.00	83	166
May-20	May-19	17.10	20.00	2.90	366	1061.4
Jun-20	Jun-19	19.57	20.78	1.21	3	3.63
Jul-20	Jul-19	16.50	19.50	3.00	245	735
Aug-20	Aug-19	16.90	22.00	5.10	273	1392.3
Sep-20	Sep-19	17.72	18.70	0.98	0	0
Oct-20	Oct-19	17.50	20.10	2.60	91	236.6
Nov-20	Nov-19	20.10	18.70	−1.40	440	−616
Dec-20	Dec-19	18.21	19.10	0.89	9	8.01
Jan-21	Jan-20	17.20	20.10	2.90	235	681.5
Feb-21	Feb-20	17.30	19.70	2.40	139	333.6
Mar-21	Mar-19	17.11	20.32	3.21	64	205.44
Apr-21	Apr-19	17.20	20.10	2.90	148	429.2
May-21	May-19	17.10	19.80	2.70	335	904.5
Jun-21	Jun-19	19.57	20.62	1.05	47	49.35
Total					2587	5725.69

Discussion

Hypotheses support

The hypothesis was supported for only one of the four categories. For investigations, once in-person learning resumed, there were no more significant declines in investigations. For CAN hotline calls, there was still a statistically significant decrease after schools resumed in-person. For both foster care entries and exits, there was actually a larger statistically significant decrease once in-person learning resumed in Florida.

In relation to decreased hotline calls, prior to the pandemic, the three largest sources of CAN reports were schools, law enforcement, and medical personnel (U.S. Children’s Bureau, 2021). Data from California shows that while the proportion of CAN reports from schools dropped by 50.0% from the 12 months from April 2020 to March 2021 from the prior 12-month period (April 2019 to March 2020), CAN reports from medical professionals increased by 13.6% and law enforcement increased by 27.1% in the same time period (Webster et al., 2022).

Even though there were higher proportions on CAN reports from medical professionals and law enforcement increased, the number of CAN reports from both of these professions decreased in the same time period. For example, in California, the number of CAN reports from medical professionals and law enforcement officers decreased by 11.3% and 0.8%, respectively (Webster et al., 2022). The proportional increase among medication professionals is due to the fact that medical professionals, like physicians play a central role in preventing and addressing CAN (Flaherty & Stirling, 2010; Humphreys et al., 2020) since they come to know a family's significant strengths and stressors, and will have frequent contact with children and their families during a crisis (Flaherty & Stirling, 2010) such as the pandemic. Even with an increase in CAN reports from medical professionals, COVID has had an impact on the ability of some physicians to identify CAN. Hospitals and other health settings have seen decreases in patient volumes (Cho et al., 2021; Edigin et al., 2020), less in-person consultation, and more social distancing measures (Cho et al., 2021; Edigin et al., 2020). At some medical training universities, medical students and interns had to take part in remote learning from home leading to a reduction in an adequate range of training on child abuse cases (Cho et al., 2021). This then reduces in-person physical examinations which is an essential skill that can only be mastered with years of practice (Edigin et al., 2020) that may lead to failing to identify sentinel injuries (such as minor injuries indicative of CAN) by not only medical students, but also less experienced physicians which could lead to a child being left in an abusive or neglectful environment (Whelan et al., 2021). In some cases, this failure may be due in part of to lack of training on the nuances of multi-faceted CAN diagnoses (Whelan et al., 2021) that was heightened during the pandemic.

There is some research to suggest similar issues around CAN identification during the COVID-19 pandemic in the law enforcement community. White et al. (2021) noted that training of police officers has historically been done primarily in-person. The shift to some of the online training has affected how police officers learn fundamental law enforcement topics (White et al., 2021). Further, under normal circumstances law enforcement receives minimal training on child abuse in comparison to their overall training. For example, in California, law enforcement receives an average of six hours of training on child abuse: four hours of standard mandated reporter training and two additional hours of specialized training for law enforcement (California Department of Social Services, 2022). This is in addition to the 4 h of training on child-related crimes in the police academy, which represents just 1.5% of the 670 h of total training that they receive in the police academy (California Commission on Peace Officer Standards and Training, 2022). This minimal training coupled with the fact that law enforcement officers had higher rates of mental health issues than other first responders (Stogner et al., 2020), has likely had an additive adverse impact on child abuse reporting.

It is surprising that not only was there a significant decrease in the number of foster care entries after the pandemic began, but that there was a larger significant decrease of foster care entries once in-person attendance resumed. Even though in-person attendance resumed for all students, not all students returned in-person. By late September 2020, only 45% of the students returned in-person in Florida (Doyle et al., 2021). Nationwide, African-American children make up the largest proportion of youth in the foster care system relative to the general population (Florida Department of Children and Families, 2022). In Florida, while African-Americans make up 16.9% of the general population in Florida (U.S. Census Bureau, 2022), they represent 27.3% of the CAN victims in the state (Florida Department of Children and Families, 2022). This may partly explain the continued significant decreases in CAN hotline calls after in-person attendance started in Florida as those youth who were at most risk for CAN, such as African-American children, were less likely to attend in-person even once schools in Florida reopened for in-person learning. African-American parents were less likely to send their children back to in-person learning due to potential fears of COVID-19 (Chua et al., 2021). African-American parents were also less likely than other racial groups to have wanted schools to reopen in person in Fall 2020, and were more likely than other racial groups to be concerned that other children would not follow COVID safety protocols, that their children could contract COVID while at school, and could bring home COVID (Gilbert et al., 2020). There is research to support the perception among African Americans that COVID-19 causes more health concerns. Research has shown that African-Americans are 2.8 times more likely than Whites to be hospitalized and 2.0 times as likely to die from COVID-19 than Whites (Tai et al., 2021).

For decreased foster care exits, this is the first study to provide statistical evidence that there was a significant decrease in foster care exits once the pandemic began. The media has reported that nationally, some 22,600 fewer YRFC left foster care from March 2020 to December 2020 than the same period one year prior (Ho & Fassett, 2021). They suggested that this was in part due the courts' decreased ability to hold child welfare hearings because of lockdowns. This is supported by a survey of child welfare professionals and caregivers (Whitt-Woosley et al., 2022), as well as the Courts (Pisani-Jacques, 2020). However, it is also surprising that there was a larger decrease in the number of foster care exits once in-person school attendance resumed which could have been further heightened by loss of foster care beds (Whitt-Woosley et al., 2022), and continued delays or suspensions of court-ordered services (Pisani-Jacques, 2020). In fact, this has led to an average of 2.2 months of longer stays in foster care. While 2.2 months extra in care does not seem long, there are adverse implications for those who stay longer in foster care. Those who stay in foster care longer are more likely to have more medical issues (Sullivan & van Zyl, 2008) and are more likely to re-enter the foster care system (Lee et al., 2012).

There is a very nuanced point about the lifetime economic cost of $633.7 million of the estimated 2338 children who were missed for CAN and not removed from foster care. There are also more costs (i.e., economic, developmental, social, behavioral, emotional, etc.) associated with youth who remain home with what would have been substantiated CAN and no foster care services. In the most extreme case, it might even lead to more deaths. Chadwick et al. (2010) noted that there are up to 14 times as many deaths related to inflicted injuries as there are non-inflicted injuries. With an approximate national child fatality rate of 7.2 per 1000 children removed into foster care (U.S. Children’s Bureau, 2021), this could mean that an estimated 17 of the 2338 children who were missed for CAN and not removed from foster care might even die because they were not removed.

Policy implications

While there may have been reductions in CAN reports and investigations, it is unlikely that these reductions represent a drop in actual CAN (Whelan et al., 2021). In fact, research has shown that there were increases in the number and proportion of CAN. For example, there was an increase in the incidence of CAN at emergency rooms and hospitals (Sharma et al., 2021), as well as an increase in the proportion of CAN after the pandemic began (Kovler et al., 2020). Further, the there was no change in severity in CAN that were seen in the hospitals. (Kaiser et al., 2021).

Research has shown that those coming into foster care are less likely to have preventative services to begin with (Sanchez et al., 2010). Up to 95% of YRFC have at least one chronic physical health issue, more than half have multiple health issues (Sanchez et al., 2010), and they have significant increased risks of serious mental health disorders, including bipolar disorder, post-traumatic stress disorder, and schizophrenia (Heim et al., 2010), and are much more at risk of being given too many psychotropic medications at the same time at too high of a level at too young of an age (Nguyen, 2021b). Almost a third of the psychotropic medications used in foster care have not been approved for use in children by the Federal Drug Administration (Nguyen, 2021b), and YRFC under the age of 12 months were five times as likely as non-YRFC to be given psychotropic medications (U.S. Government Accountability Office, 2011). This is why back in 2007, the American Academy of Pediatrics along with three other primary care associations advocated for a set of joint principles using a medical home model (Sanchez et al., 2010) for vulnerable child populations which is important for YRFC. These principles included having a personal physician providing coordinated whole person integrated care through one health care system (Sanchez et al., 2010). These principles are true today, and are especially important given that non-urgent medical care has been postponed or cancelled because of the pandemic (Edigin et al., 2020).

Another policy implication is that there needs to be an interoperable system with real-time data access for child welfare jurisdictions based on some elements of predictive analytics. Russell (2015) notes that “…Predictive analytics can include a broad set of statistical and analytical tools that identify trends, relationships, and patterns within data that can be used to predict a future event or behavior.” (p. 183). There have been a number of successful partnerships between universities and public agencies on predictive analytics, and 11 states have deployed predictive analytics tools including the Alleghany [County, Pennsylvania, U.S.A.] Family Planning Screening Tool originally developed by researchers at the University of Queensland (New Zealand) and the University of Southern California (U.S.A.) (Loudenback 2022). In fact, children maltreated and classified by this predictive analytics tool have been found to have increased foster care placements, emergency room visits and hospitalizations for related injuries (Vaithianathan et al., 2020). Unfortunately, while the best predictive analytics model using the most advance statistical methods can only offer what might happen, it takes people with expertise to determine what to do with the information (Russell, 2015).

Limitations

There are several major limitations of this study. The first major limitation is that even the few state databases that have aggregate data do not share enough detail. For example, California's child welfare data dashboard reports quarterly data (Webster et al., 2022), Florida's child welfare data dashboard does not break down where the reports come from (Florida Department of Children and Families, 2022), and Wisconsin's data dashboard combines some reporting types, specifically medical and mental health as one category (Wisconsin Department of Children and Families, 2022). It is also difficult to determine trends for specific types of child abuse allegations because of the lack of detail. For example, it would be difficult to examine if there was an increase in number of hotline calls related to educational neglect during the time of remote learning. Further, the Federal Government's Child Maltreatment Report is 16 months old as soon as it is published in January of each year (U.S. Children’s Bureau, 2021). Currently, there is also no way to utilize interoperable systems as they do not exist between child welfare systems and electronic health records, even though such interoperable systems could be developed (Nguyen, 2014).

Another limitation was that this study did not address seasonal fluctuations in CAN hotline calls such as during the holiday seasons in the U.S. (e.g., November–December). Researchers have examined the brief period following the start of the pandemic (i.e., March 2020 to May 2020) and have found statistically-significant decreases in CAN reports (Rapoport et al., 2021). As the COVID situation becomes more of an endemic issue in which we are able to better delineate the effects of seasonality, it will be very important to examine CAN reporting in the different seasons.

Possible innovations

Most of the current examples related to predictive analytics have focused on public agency-academic/research partnerships built upon strong relationships between leaders and executives in public agencies and the faculty and researchers from the universities and other research entities. These partnerships have a tendency to be hampered when information from one entity has to be screened/vetted or continually transferred between entities. The partnerships may be further compromised if there are not enough experts in the public agency to understand the results of the research from academic/research entity, or there are not enough experts in the academic/research entity to understand the nuances of the impact of daily public agency operations. In essence, the interaction between the practitioners and the academics/researchers is critical as the data tools alone lack the practice insights and contextual knowledge to be able to provide truly valuable knowledge (Russell, 2015).

We need to build the human resources within public child welfare organizations in senior positions that have the expertise to lead efforts to analyze and interpret predictive analytics based on sophisticated statistical models, coordinate short-term efforts to respond to such analytics, and lay the foundation to influence and direct major policy and practice change. The best way to do this is to develop opportunities for public child welfare staff to pursue doctorates (including those in social work) or to recruit those who have such advance degrees in academia/research to work in the public agency settings. In the social work field, there are two types of doctorates, the research doctorate (i.e., Ph.D.), and the professional doctorate (i.e., the Doctor of Social Work degree or D.S.W.) (CSWE/GADE, 2021). The research doctorate is more focused on research and careers are usually in the research or academic arenas while about half of the D.S.W. programs are focused on clinical practice while the other half tend to focus on community practice and teaching (CSWE/GADE, 2021).

We should model a program like the Title IV-E stipend for Bachelor's of Social Work and Master's of Social Work students to pursue careers in public child welfare (CalSWEC, 2022) in which for each year of the stipend, students work a year in public child welfare. With the innovation, but with a focus on doctoral-level pursuits, Title IV-E fellowships would be developed for doctoral students who want to pursue future careers focused on predictive analytics, applied research, quality improvement, and evaluation in child welfare organizations. The focus would be on developing leaders and executives who have the strong backgrounds in both the research and the practice approaches of doctoral-level learning. Child welfare organizations could then develop opportunities for such leaders and executives to take lead roles within the organization on predictive analytics, applied research, quality improvement, and evaluation or even members of agency executive teams.

Conclusion

Once in-person learning resumed in all schools in Florida, there continued to be statistically-significant decreases in CAN hotline calls, and larger statistically-significant decreases in foster care entries and exits. This meant that beginning November 2020, there were an estimated 4769 CAN hotline calls, 844 fewer foster care entries, and 1370 fewer foster care exits than the year before the pandemic began. The COVID-19 Omicron variant is more transmissible than previous variants which has led to a doubling of positive COVID-19 cases every three days in some parts of the world (Mahase, 2021). Since the first COVID-19 Omincron variant was detected in Florida in early December 2021, this variant is more infectious than the Delta variant with the potential of infecting most of Florida's population, especially among unvaccinated individuals (Hladish et al., 2022). As even more variants are identified, this trend will likely continue to lead to statistically significant decreases in CAN hotline calls, foster care entries, and exits even with the resumption of in-person learning in all Florida school districts.

The issues noted thus far, as well as the limitations highlight the need to go beyond an a posteriori analysis of the trends of CAN hotline calls, CAN investigations, foster care entries and exits during the COVID pandemic. We need to be able to have an interoperable data system with real-time data that has the specificity to provide meaningful decision-making. Further, we need leaders and administrators with advance degrees like the Ph.D. and D.S.W. in order effectively address the future impact of disasters like the COVID-19 pandemic.