Advertisement

Enhancing Quality Measurement With Clinical Information: A Use Case of Body Mass Index Change Among Children Taking Second Generation Antipsychotics

      Abstract

      Objective

      We sought to examine the extent to which body mass index (BMI) was available in electronic health records for Florida Medicaid recipients aged 5 to 18 years taking Second-Generation Antipsychotics (SGAP). We also sought to illustrate how clinical data can be used to identify children most at-risk for SGAP-induced weight gain, which cannot be done using process-focused measures.

      Methods

      Electronic health record (EHR) data and Medicaid claims were linked from 2013 to 2019. We quantified sociodemographic differences between children with and without pre- and post-BMI values. We developed a linear regression model of post-BMI to examine pre-post changes in BMI among 4 groups: 1) BH/SGAP+ children had behavioral health conditions and were taking SGAP; 2) BH/SGAP- children had behavioral health conditions without taking SGAP; 3) children with asthma; and 4) healthy children.

      Results

      Of 363,360 EHR-Medicaid linked children, 18,726 were BH/SGAP+. Roughly 4% of linked children and 8% of BH/SGAP+ children had both pre and post values of BMI required to assess quality of SGAP monitoring. The percentage varied with gender and race-ethnicity. The R2 for the regression model with all predictors was 0.865. Pre-post change in BMI differed significantly (P < .0001) among the groups, with more BMI gain among those taking SGAP, particularly those with higher baseline BMI.

      Conclusion

      Meeting the 2030 Centers for Medicare and Medicaid Services goal of digital monitoring of quality of care will require continuing expansion of clinical encounter data capture to provide the data needed for digital quality monitoring. Using linked EHR and claims data allows identifying children at higher risk for SGAP-induced weight gain.

      Keywords

      What's New
      Incorporating clinical information into quality measurement holds promise for measuring both processes and outcomes of care. Missing clinical information disproportionately affects different subgroups of children, potentially contributing to inequities in assessing the quality of care.
      As many as 10% of children in Medicaid overall and 22% in foster care take antipsychotic medication.
      • Crystal S
      • Mackie T
      • Fenton MC
      • et al.
      Rapid growth of antipsychotic prescriptions for children who are publicly insured has ceased, but concerns remain.
      ,
      • Keast SL
      • Tidmore LM
      • Shropshire D
      • et al.
      Characterization of chronic multiclass psychotropic polypharmacy and psychotherapy in foster care youth in a state Medicaid population.
      Antipsychotic use among children peaked around 2005 and has declined since then due to state Medicaid Program measures.
      • Leckman-Westin E
      • Finnerty M
      • Scholle SH
      • et al.
      Differences in Medicaid antipsychotic medication measures among children with SSI, foster care, and income-based aid.
      Despite the decline in use, children continue to be placed on the medications and require monitoring to ensure high-quality care. The Pediatric Quality Measures Program
      • Mistry KB
      • Chesley F
      • LLanos K
      • et al.
      Advancing children's health care and outcomes through the pediatric quality measures program.
      and the National Collaborative for Innovation in Quality Measurement led the development of the “Safe and Judicious Use of Antipsychotics in Children” measure set.

      All PQMP Measures. Available at: http://www.ahrq.gov/pqmp/measures/all-pqmp-measures.html. Accessed April 14, 2021.

      ,

      National Collaborative for Innovation in Quality Measurement. NCQA. Available at: https://www.ncqa.org/hedis/reports-and-research/national-collaborative-for-innovation-in-quality-measurement/. Accessed April 27, 2021.

      Metabolic monitoring is a component of that measure set and includes whether children received glucose and lipid testing annually. However, the measure does not incorporate clinical information thereby limiting the ability to improve children's health outcomes, as opposed to processes alone.
      Body mass index (BMI) is a particularly important clinical outcome because weight gain is one of the most common side effects of second-generation antipsychotics (SGAPs), with up to 80% of children gaining weight, depending on the medication used.
      • Dayabandara M
      • Hanwella R
      • Ratnatunga S
      • et al.
      Antipsychotic-associated weight gain: management strategies and impact on treatment adherence.
      SGAP-induced weight gain is observed more in children than in adults and may further contribute to long-term psychological distress and social withdrawal throughout early adulthood.
      • Correll CU
      • Carlson HE.
      Endocrine and metabolic adverse effects of psychotropic medications in children and adolescents.
      Clinical information can inform quality improvement in 3 ways: 1) identification of children at greatest risk of weight change for enhanced monitoring and early intervention, 2) implementation of lifestyle interventions tailored to children's unique needs,
      • Alia KA
      • Wilson DK
      • McDaniel T
      • et al.
      Development of an innovative process evaluation approach for the Families Improving Together (FIT) for weight loss trial in African American adolescents.
      • Janicke DM
      • Lim CS
      • Mathews AE
      • et al.
      The Community-based Healthy-lifestyle Intervention for Rural Preschools (CHIRP) study: design and methods.
      • Janicke DM
      • Sallinen BJ
      • Perri MG
      • et al.
      Comparison of program costs for parent-only and family-based interventions for pediatric obesity in medically underserved rural settings.
      and 3) development of provider capacity to intervene with children experiencing weight gain.
      The Centers for Medicare and Medicaid Services (CMS) is transitioning to digital quality measures (dQMs) that use data from electronic health records (EHRs), registries, and other sources to reduce reporting burden while also providing access to clinical information that can support evaluation and care improvement.

      Meaningful Measures 2.0: Moving from Measure Reduction to Modernization | CMS. Available at:https://www.cms.gov/meaningful-measures-20-moving-measure-reduction-modernization. Accessed October 31, 2021.

      dQMs provide more refined information to assess quality of care and to implement evidence-based interventions including enhanced screening.

      Hamlin B, O'Kane M, Barr M, et al. COVID-19 Underscores The Need For Digital Quality Measurement | Health Affairs Blog. Available at: https://www.healthaffairs.org/do/10.1377/hblog20200824.987845/full/. Accessed October 31, 2021.

      Further, dQMs have the potential to facilitate rapid-cycle feedback and learning because the data are captured more readily from EHRs in contrast to traditional quality measures that rely on health care claims and/or medical record abstraction. dQMs also have the potential to inform a learning measurement system, where the measures themselves are rapidly refined based on user experience and lessons learned.

      Hamlin B, O'Kane M, Barr M, et al. COVID-19 Underscores The Need For Digital Quality Measurement | Health Affairs Blog. Available at: https://www.healthaffairs.org/do/10.1377/hblog20200824.987845/full/. Accessed October 31, 2021.

      While CMS's goal is to have all quality measures digital by 2030, several key considerations must be addressed to spur dQM development and use. The availability and accuracy of clinical data from EHRs are central to the transition to dQMs, yet are potentially a hurdle to the advancement of a digital quality ecosystem. In this study, we examined the extent to which clinical information about BMI or height and weight to calculate BMI, were available in EHRs for children taking SGAPs. A single record of BMI or height and weight is not enough. Assessing weight gain requires 2 or more measurements spaced sufficiently apart in time. Moreover, the requirement of 2 or more measures at different time points is consistent with CMS guidelines for Medicaid Program performance improvement projects. Therefore, we compared the sociodemographic characteristics of children with 2 or more BMI values to those with one or no measures to identify potential biases in the data. Further, within the subset of children with two BMI measures, we examined how the clinical data can be used to identify children most at-risk for weight gain, which cannot be done using process-focused measures.

      Methods

      Data Sources

      We used Florida Medicaid Program claims linked to the children's EHR data contained in the OneFlorida Data Trust. The OneFlorida Data Trust is a centralized data repository containing EHR data for 17 million Floridians from 10 health care systems from 2012 to the present and refreshed quarterly.
      • Hogan WR
      • Shenkman EA
      • Robinson T
      • et al.
      The OneFlorida Data Trust: a centralized, translational research data infrastructure of statewide scope.
      Together the health systems form the OneFlorida Clinical Research Consortium, one of nine Patient-Centered Outcomes Research Institute-funded sites nationally. A data use agreement between the University of Florida and Florida Medicaid allows for the data linkage. The linked data are Health Insurance Portability and Accountability Act-limited data sets, which restrict protected health information to dates (eg, birthdates) and location (5 or 9-digit zip code level, which allows for geocoding). The health system partners submit the data to the Data Trust and it is harmonized using the Patient-Centered Outcomes Research Institute's Common Data Model.

      Patient Centered Outcomes Research Institute. PCORnet Common Data Model. Available at:https://pcornet.org/wp-content/uploads/2019/09/PCORnet-Common-Data-Model-v51-2019_09_12.pdf. Accessed September 20, 2020.

      The data undergo a quality characterization every six months through the PCORnet Coordinating Center and must meet pre-defined standards to be approved for study use.
      Variables were obtained from children's EHRs and Medicaid claims and enrollment data. Claims data were used to identify medication use (name, type, fill date, number of supply days), diagnosis (codes and dates), 5- or 9-digit zip codes, and enrollment (enrollment status and month). The children's health status was computed from claims data using the 3M Clinical Risk Groups.

      3M. 3M™ Clinical Risk Groups (CRG) Classification System Methodology Overview. Available at: https://apps.3mhis.com/docs/Groupers/Clinical_Risk_Grouping_CRG/methodology_overview/grp401_crg_v2.2_meth_overview.pdf. Accessed December 14, 2021.

      Enrollment files were used to identify the children's race and ethnicity and age and to calculate their social vulnerability index and Rural-Urban Continuum Codes using 5- or 9-digit zip codes. BMI and obese status were computed using the height, weight and measure date from EHRs.

      Study Design and Inclusion Criteria

      This is a cohort study in Medicaid-insured children and adolescents with behavioral health conditions using claims and EHR data from 2013 through 2019. Children 5 to 18 years old who met the following eligibility criteria were included to 1) examine the extent to which BMI or height and weight were available in EHR data and 2) quantify the characteristics of children most at risk for weight gain to illustrate how the information could be used for quality improvement planning.
      We included 4 study groups: an exposure group, a control group and two comparison groups. The exposure group contained children diagnosed with at least one behavioral health condition and subsequently taking SGAP (BH/SGAP+ group). The control group contained children who had at least one behavioral health condition without any SGAP (BH/SGAP- group). Because the use of SGAP was not assigned randomly, we included two additional comparison groups to account for the expected increase in weight as children grow: 1) children who had no behavioral health conditions but were diagnosed with asthma, a common chronic physical condition (Asthma group) and 2) children determined as healthy according to 3M Clinical Risk Groups (Healthy group).

      3M. 3M™ Clinical Risk Groups (CRG) Classification System Methodology Overview. Available at: https://apps.3mhis.com/docs/Groupers/Clinical_Risk_Grouping_CRG/methodology_overview/grp401_crg_v2.2_meth_overview.pdf. Accessed December 14, 2021.

      For each group, we identified the numbers of children with BMI and/or height and weight information. We also compared the sociodemographic characteristics between children with and without both pre- and post-BMIs to quantify potential biases that may be present when examining the clinical outcome.
      To quantify the characteristics of children most at risk for weight gain, we examined the presence and duration of SGAP use using medication fill date and supply days from Medicaid claims dispensing data. Children were SGAP+ if they had at least one month of drug use. SGAP prescriptions were identified using National Drug Codes.

      U.S. Food and Drug Administration. National Drug Code Directory. Available at: https://www.accessdata.fda.gov/scripts/cder/ndc/index.cfm. Accessed April 22, 2021.

      Behavioral health conditions and asthma were identified using the International Statistical Classification of Diseases 9th and 10th Revision. The child needed to have one inpatient or two outpatient encounters using E & M physician visit code visits with the behavioral health or asthma diagnosis. We used the earliest observed behavioral condition diagnosis as the index diagnosis if a child had multiple behavioral conditions.
      Because the BMI data were from EHRs without a pre-specified schedule for data collection, we selected an index date as the baseline for each child. The index date was defined as the first date observed for SGAP dispensing for the BH/SGAP+ group, the behavioral health condition diagnosis for the BH/SGAP-group, the asthma diagnosis for the Asthma group, or January 1 of the year with 3M Clinical Risk Group status determined as healthy for the Healthy group. BMI at baseline is defined as the BMI measurement 0 to 6 months before the index date. The children also had to have a post-BMI measured 12 ± 3 months after the index date. One single year of the outcome data for each participant was included in the analysis. To simplify the study, we did not include multiple years of post-BMIs in the analysis since the number of available BMI measurements and the time intervals between BMI or height and weight measurements varied among children.
      Although we do not have medication information prior to Medicaid enrollment, more than 75% of the children in the BH/SGAP+ group had at least 12 months between the enrollment date and the date of their first SGAP use, which can be considered a washout period to establish incident users.

      Outcomes

      There are two outcomes of interest: 1) variations in the sociodemographic characteristics of children with and without two recorded BMIs, and 2) expected change in BMI 12 ± 3 months after initiating SGAP associated with sociodemographic characteristics and baseline health.
      For children and adolescents, BMI is influenced by normal growth and development and body fat accumulation. Change in BMI is also age- and sex-dependent. We chose BMI rather than age- and sex-adjusted BMI z-score because 1) we were interested in the subgroup difference between males and females in weight gain and 2) BMI z-scores have limitations in tracking individuals with severe obesity.
      • Freedman DS
      • Berenson GS.
      Tracking of BMI z scores for severe obesity.
      ,
      • Freedman DS
      • Butte NF
      • Taveras EM
      • et al.
      BMI z-scores are a poor indicator of adiposity among 2- to 19-year-olds with very high BMIs, NHANES 1999-2000 to 2013-2014.
      We used the Centers for Disease Control method to detect and remove biologically implausible values for data cleaning.

      SAS Program (ages 0 to < 20 years) | Resources | Growth Chart Training | Nutrition | DNPAO | CDC. 2019. Available at:https://www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas.htm. Accessed March 12, 2021.

      We assessed the data quality using a univariate correlation between BMI at baseline and post-BMI after data cleaning.

      Participant Characteristics

      Participant characteristics used as predictor variables included age, sex, race-ethnicity, rurality of residence, social vulnerability index, baseline BMI, and study groups. Race-ethnicity was grouped as Hispanic, non-Hispanic (NH)-White, NH-Black, NH-Other, and Unknown. Rurality was determined by aggregating Rural-Urban Continuum Codes with 1 as urban, 2 to 3 as metro, and 4 to 9 as rural area.

      USDA ERS - Rural-Urban Continuum Codes. Available at: https://www.ers.usda.gov/data-products/rural-urban-continuum-codes.aspx. Accessed March 12, 2021.

      The social vulnerability index is a composite measure of social vulnerability at the census tract level where census tracts are ranked using 15 social factors (eg, socioeconomic status, household composition, housing, transportation).

      Centers for Disease Control. Social Vulnerability Index. Available at: https://svi.cdc.gov/factsheet.html. Accessed August 27, 2020.

      Statistical Analyses

      To quantify differences in sociodemographic characteristics for children with or without both pre- and post-BMIs in the EHR, we used the nonparametric Wilcoxon 2-sample test for the continuous variable of age, and the chi-square test for the categorical variables. For the use case examining BMI changes, we fit general linear regression models with post-BMI as the outcome. Both linear and quadratic terms of age to control for the nonlinear relationship between BMI and age within the study cohort. Age was centered by the sample mean of 11.3 years. Interaction terms that caused collinearity were not included. The full model included all the primary predictors and 2-way interactions between study group and sex, race-ethnicity, baseline BMI, rurality, social vulnerability index. We also tested 2-way interactions between sex by race-ethnicity, sex by baseline BMI, and race-ethnicity by baseline BMI. Only significant interactions remained in a final model.
      We conducted a residual analysis to check if model assumptions were satisfied. Regression coefficients and their standard errors, and P values of all the predictors were reported. A nominal level of P value ≤ .05 was used for statistical significance. Changes in R2 were also reported as the measures of association between the outcome and the predictors to assess the importance of one or more variables in predicting post-BMI. All statistical analyses were performed in SAS Version 9.4 (SAS Institute, Cary, NC).

      Results

      Sociodemographic characteristics between children with and without pre- and post-BMI records were compared (Table 1). A total of 363,360 children were identified from the Data Trust who were in the BH/SGAP+, BH/SGAP-, Asthma or Healthy groups. Of these children, only 4.2% had both pre- and post-BMI records in the EHR. Among BH/SGAP+ children, the percentage was 7.6%. A higher percentage of females relative to males and a lower percentage of NH-Black children had both pre- and post-BMI records (P < .0001). A higher percentage of two recorded BMIs was also observed among children in metro and rural areas (P < .0001). No significant differences were observed based on socioeconomic vulnerability.
      Table 1Sociodemographic Characteristics of Participants With or Without Both Pre- and Post-BMI Records by Study Group at Baseline
      Participant characteristics were compared between children with both pre- and post-BMIs (2+ BMI) and children without both BMI measures (0–1 BMI) within each study group and in the whole study population. Continuous variable age was compared using a non-parametric Wilcoxon two-sample test, and categorical variables were compared using the chi-square test. All comparisons are statistically significant (P value < .0001), except for the social vulnerability index in BH/SGAP+ group (P value = .18).
      Study Group
      BH/SGAP+BH/SGAP-AsthmaHealthyAll
      Variable0–1 BMI (n = 17,309)2+ BMI (n = 1417)0-1 BMI (n = 125,979)2+ BMI (n = 8597)0-1 BMI (n = 21,970)2+ BMI (n = 1763)0-1 BMI (n = 182,860)2+ BMI (n = 3465)0-1 BMI (n = 348,118)2+ BMI (n = 15,242)
      Age (mean ± SD)12.4 ± 3.512.3 ± 3.610.7 ± 3.811.3 ± 3.910.0 ± 3.610.2 ± 3.711.5 ± 3.811.8 ± 4.011.2 ± 3.811.3 ± 3.9
      Female (%)40.9%46.1%47.2%52.8%47.4%51.2%53.9%58.2%50.4%53.2%
      Race/ethnicity
       NH-White (%)33.5%30.8%28.8%30.1%17.2%14.4%22.5%19.6%25.0%25.9%
       NH-Black (%)18.8%14.2%25.4%23.9%31.2%32.6%30.6%29.9%28.2%25.4%
       Hispanic (%)16.1%19.6%31.0%29.2%41.2%42.6%36.9%39.0%34.0%32.1%
       NH-Other (%)4.7%4.6%6.9%7.1%8.9%7.8%9.1%10.1%8.0%7.6%
       Unknown (%)26.9%30.8%7.9%9.7%1.5%2.6%1.0%1.4%4.8%8.9%
      Rurality
       Rural (4–9) (%)4.7%6.9%4.6%5.5%3.0%2.3%3.3%2.2%3.8%4.5%
       Metro (2–3) (%)30.7%27.0%21.1%24.6%17.8%13.6%17.4%15.7%19.4%21.5%
       Urban (1) (%)55.1%57.4%65.9%61.9%71.4%78.0%71.5%74.6%68.7%66.2%
       Undetermined (%)9.5%8.7%8.4%7.9%7.9%6.1%7.9%7.5%8.1%7.7%
      Social vulnerability index
       1st quartile (%) (least vulnerable)9.8%9.6%9.0%9.3%6.9%7.5%8.9%9.2%8.9%9.1%
       2nd quartile (%)18.1%19.1%17.7%18.7%16.2%15.4%17.7%17.7%17.6%18.1%
       3rd quartile (%)25.1%24.8%25.2%26.2%25.5%26.0%25.9%26.4%25.6%26.1%
       4th quartile (%) (most vulnerable)37.5%37.8%39.8%38.0%43.4%44.9%39.7%39.2%39.8%39.0%
       Undetermined (%)9.5%8.7%8.4%7.9%7.9%6.1%7.9%7.5%8.1%7.7%
      BH indicates behavioral health; SGAP, second-generation antipsychotics; NH, non-Hispanic; and BMI, body mass index.
      low asterisk Participant characteristics were compared between children with both pre- and post-BMIs (2+ BMI) and children without both BMI measures (0–1 BMI) within each study group and in the whole study population. Continuous variable age was compared using a non-parametric Wilcoxon two-sample test, and categorical variables were compared using the chi-square test. All comparisons are statistically significant (P value < .0001), except for the social vulnerability index in BH/SGAP+ group (P value = .18).
      To examine changes in BMI, 15,242 children with both pre- and post-BMI records were included in a regression analysis. Figure 1 illustrates the sample selection process and the reason for elimination at each step. Sociodemographic characteristics for the regression analysis are presented in Table 1. The BH/SGAP+ group was 46.1% female, compared to 52.8% females in the BH/SGAP- group, 51.2% in the Asthma group, and 58.2% in the Healthy group. There were 14.2% NH-Blacks and 19.6% Hispanics in the BH/SGAP+ group, compared to 23.9% NH-Blacks and 29.2% Hispanics in the BH/SGAP- group. Further, 30.8% of children had an unknown race in BH/SGAP+ group, higher than those in BH/SGAP- group (9.7%), Asthma group (2.6%), and Healthy group (1.4%).
      Figure 1
      Figure 1Flow chart of the participants included in the analysis from linked OneFlorida Data Trust and Florida Medicaid Claims Data between January 1, 2013 and December 31, 2019.
      Higher percentages of children with behavioral health conditions (6.9% in the BH/SGAP+ group and 5.5% in the BH/SGAP- group) live in rural areas than those without behavioral health conditions (2.3% in the Asthma group and 2.2% in Healthy group). There were lower percentages of children with behavioral health conditions living in urban areas with 57.4% of children in the BH/SGAP+ group and 61.9% in the BH/SGAP- group, in contrast to 78.0% in the Asthma group and 74.6% in the Healthy group. The distributions of social vulnerability index quartiles were similar across the four groups. The baseline BMI was slightly higher in BH/SGAP+ (21.6 ± 6.1) compared to the other three groups (21.2 ± 6.1 in BH/SGAP-, 20.5 ± 5.7 in Asthma, and 21.1 ± 5.7 in the Healthy group).
      Table 2 presents the clinical characteristics of the children. A higher percentage of BH/SGAP+ children had obesity (23.0%) compared to healthy children (20.3%). Behavioral health conditions differed between children taking or not taking SGAP medications. Higher percentages of children with conditions for which SGAP had first-line indications (schizophrenia, autism and bipolar disorder) were in the BH/SGAP+ groups than in the BH/SGAP- groups. Behavioral health conditions associated with the highest SGAP use were ADHD with (12.8%) and without SGAP use were conduct disorder (19.4%), anxiety or depression (16.4%), and other mental health disorders (28%). Risperidone was the most frequently used SGAP (49.3%), followed by aripiprazole (23.1%), and quetiapine (16.4%).
      Table 2Clinical Characteristics in Children With 2 or More BMIs by Study Group at Baseline
      Study Group
      Clinical VariableBH/SGAP+ (n = 1417)BH/SGAP- (n = 8597)Asthma (n = 1763)Healthy (n = 3465)Total (n = 15,242)
      BMI at baseline21.6 ± 6.121.2 ± 6.120.5 ± 5.721.1 ± 5.721.1 ± 6.0
      Obesity (%)23.024.725.420.323.6
      BH condition (%)*
       Schizophrenia1.40.4..0.4
       Autism w/o irritability6.82.1..1.8
       Autism with irritability2.80.0..0.3
       Bipolar3.51.2..1.0
       ADHD w/o conduct disorder19.424.8..15.8
       ADHD with conduct or disruptive disorder12.82.1..2.4
       Conduct or disruptive disorder, no ADHD1.22.0..1.2
       Anxiety or depression16.423.6..14.8
       Trauma and stressor / adjustment related7.713.0..8.0
       Other mental health disorder28.030.9..20.0
       No BH condition..100.0100.034.3
      Type of antipsychotics
       Risperidone49.26....
       Olanzapine5.01....
       Quetiapine16.37....
       Aripiprazole23.08....
       Asenapine0.00....
       Clozapine0.07
       Lurasidone1.98....
       Ziprasidone4.52....
       Iloperidon0.00....
       Paliperidone0.35....
      BH indicates behavioral health; SGAP, second-generation antipsychotics; BMI, body mass index; and ADHD, attention deficit hyperactivity disorder.
      The univariate correlation coefficient was 0.928 between BMI at baseline and post-BMI measured 12 ± 3 months after the index date. In turn, R2 equals 0.9282 = 0.861, which indicates 86% post-BMI variability is predictable by baseline BMI.
      The final regression model predictors were age, age2, sex, race-ethnicity, sex × race-ethnicity, rurality, social vulnerability index, baseline BMI, baseline BMI × sex, baseline BMI × race-ethnicity, study group, study group × sex, and study group × baseline BMI (Table 3). Overall, the model fit was excellent with an R2 of 0.8654.
      Table 3Regression Model Predicting BMI 12 ± 3 Months After Baseline
      VariableRegression CoefficientStandard ErrorP ValueΔR2
      Intercept21.790.10<.0001
      Age in year
      Age was centered by subtracting its mean 11.3 years.
      0.050.01<.00010.2647
      Age2-0.010.00<.00010.0027
      Sex (ref = female)-0.070.11<.00010.0041
      Race-ethnicity (ref = White)<.00010.0019
       Black0.130.08
       Hispanic0.150.07
       Other0.120.11
       Unknown0.380.11
      Sex × race-ethnicity<.00010.0015
       Male × Black-0.190.11
       Male × Hispanic0.010.10
       Male × other-0.080.16
       Male × unknown-0.330.15
      Rurality (ref = urban)
       Metro0.060.05<.00010.0002
       Rural-0.010.09<.00010.0016
      Social vulnerability index (ref = 1st quartile)<.00010.0018
       2nd quartile0.090.08
       3rd quartile0.110.07
       4th quartile (most vulnerable)0.180.07
       Undetermined0.180.09
      Baseline BMI0.960.01<.00010.5842
      Baseline BMI × Male-0.020.01.00170.0001
      Baseline BMI × Race-ethnicity.02450.0001
       Baseline BMI × Black0.010.01
       Baseline BMI × Hispanic-0.010.01
       Baseline BMI × other-0.010.01
       Baseline BMI × unknown0.020.01
      Group (ref = healthy)
      BH/SGAP+: children having behavioral conditions and taking second-generation antipsychotics; BH/SGAP-: children having behavioral conditions without second-generation antipsychotics; Asthma: children with asthma; Healthy: healthy children.
      <.00010.0021
       BH/SGAP+1.050.11
       BH/SGAP-0.020.06
       Asthma-0.100.09
      Sex × group.00130.0001
       Male × BH+/SGAP+-0.130.16
       Male × BH/SGAP--0.120.10
       Male × asthma0.330.14
      Baseline BMI × group.00010.0002
       Baseline BMI × BH/SGAP+0.060.01
       Baseline BMI × BH/SGAP-0.010.01
       Baseline BMI × asthma0.000.01
      BH indicates behavioral health; SGAP, second-generation antipsychotics; and BMI, body mass index.
      low asterisk Age was centered by subtracting its mean 11.3 years.
      BH/SGAP+: children having behavioral conditions and taking second-generation antipsychotics; BH/SGAP-: children having behavioral conditions without second-generation antipsychotics; Asthma: children with asthma; Healthy: healthy children.
      In the final model, the results for primary predictors were all statistically significant. The post-BMI measure was strongly associated with social vulnerability index (ΔR2 = 0.0018), rurality (ΔR2 = 0.0002 + 0.0016 = 0.0018), race-ethnicity (ΔR2 = 0.0019), sex (ΔR2 = 0.0041), baseline BMI (ΔR2 = 0.5842), and age (ΔR2 = 0.2674 including both the linear and the quadratic terms of age). Approximately 27% of the variation in post-BMI is explained by age and 58% is explained by baseline BMI (adjusted for age, sex, race-ethnicity, interaction between sex and race-ethnicity, rurality, and social vulnerability index). There were small but statistically significant associations between post-BMI and race-ethnicity, rurality and social vulnerability index. The large sample size in this study allowed detecting small clinical effects, with all P values less than .0001 for the main effects of the predictors.
      Interactions were examined in detail. Females had a higher predicted post-BMI than males in all study groups except Asthma. The predicted post-BMI was 22.7 in males in BH/SGAP+, and 21.7 in males in BH/SGAP-, showing a 1.02 greater increase in BMI in BH/SGAP+ compared to BH/SGAP- in males. The significant interaction of baseline BMI with the study group indicates that the slope of the regression line predicting post-BMI from baseline BMI differs by study group. The slope is 1.008 ± 0.011 in BH/SGAP+ group, which is higher than the slopes in the control and the comparison groups (Fig. 2). As expected, children taking SGAP had an increase in BMI gain as opposed to those not taking the medications. The increase in BMI was even higher in children who had higher BMI at baseline for all cohorts. This finding is particularly important for children taking SGAP who have a greater risk for BMI increases across time relative to children not taking SGAP.
      Figure 2
      Figure 2Regression plots of BMI at baseline versus post-BMI by study group.

      Discussion

      The “Safe and Judicious Use of Antipsychotics in Children and Adolescents” measurement set includes assessing metabolic monitoring, that is, whether glucose and lipid tests were conducted. Clinical information about the children's actual laboratory results, weight and height (or BMI) is not captured. In our study, we focused on assessing repeated measures of BMI for three primary reasons. First, BMI records should be available in the EHR data for all children. Second, childhood obesity is associated with an increased risk of cardiovascular disease, type 2 diabetes, anxiety and depression.
      • Lanigan J.
      Prevention of overweight and obesity in early life.
      ,
      • Romero-Pérez EM
      • González-Bernal JJ
      • Soto-Cámara R
      • et al.
      Influence of a physical exercise program in the anxiety and depression in children with obesity.
      Third, there is some evidence that overweight and obesity contribute to poorer outcomes among children with serious mental illnesses such as bipolar disorder.
      • McIntyre RS
      • Mansur RB
      • Lee Y
      • et al.
      Adverse effects of obesity on cognitive functions in individuals at ultra high risk for bipolar disorder: results from the global mood and brain science initiative.
      Digital quality measures, with their use of EHR data, have the potential to guide quality improvement efforts toward addressing outcomes of care. However, careful attention must be paid to data quality and to biases that may be present in the data. In our case, children who are male or NH-Black may not be accurately represented in dQMs focused on BMI. We do not know if the missing data are related to a lack of obtaining or recording BMI or height and weight. This is a finding that requires further exploration of the potential causes to improve care and data quality.
      Using linked Medicaid claims and EHR data, for those children with at least two BMI records, we assessed the impact of multiple variables on weight gain in children taking SGAP. At baseline, 17.3% of children included in this study overall had a BMI indicative of overweight and 23.6% had a BMI indicative of obesity for a total of 41% of children with overweight or obesity. Children taking SGAP had a significant increase in BMI compared to children without SGAP, which is consistent with the literature.
      • Dayabandara M
      • Hanwella R
      • Ratnatunga S
      • et al.
      Antipsychotic-associated weight gain: management strategies and impact on treatment adherence.
      Health and sociodemographic characteristics were positively associated with weight gain. For example, females and those with higher BMIs at baseline were most at-risk for greater weight gain when taking SGAPs pointing to opportunities for early intervention with these subgroups, in particular.
      A key feature of this study was the use of EHR data to examine clinical outcomes for a large group of children. However, we found that among the 18,726 children taking SGAPs only 7.6% had two BMIs recorded at sufficient intervals to examine change across time. This raises the importance of addressing the measurement and documentation of BMI in the EHR as a quality of care concern.
      Quality of care measurement has historically relied on process measures and health care claims data. However, attempts are being made to better measure clinical outcomes by using information from EHRs.
      • Burstin H
      • Leatherman S
      • Goldmann D.
      The evolution of healthcare quality measurement in the United States.
      While the OneFlorida Data Trust and the partnership with Florida Medicaid is an example of linking EHR and claims data to better understand clinical outcomes, more work remains to be done to ensure key measures of children's health are captured, such as height, weight and BMI. This is especially important in light of our finding that higher percentages of children who were males and NH-Blacks, for example, were excluded from the regression analyses requiring two or more BMI records, relative to females and white children.
      There are limitations to the study. First, we are relying on EHR data from ten health systems in Florida, which may not be representative of the quality of EHR data from other providers. Second, children without SGAP have different distributions of sex, race-ethnicity, and particular behavioral disorders compared to those taking SGAP. Adding comparison groups partially assesses, but may not completely control for confounding. Third, while generally considered a drug class effect, the degree of weight gain following SGAP exposure varies across individual drugs.
      • Ben Amor L.
      Antipsychotics in pediatric and adolescent patients: a review of comparative safety data.
      We did not examine the effects of specific SGAP medications. Finally, the results may not be generalizable to children who do not have the characteristics of those included in the study. While non-Hispanic white children comprised only 26% of the children in the study, underrepresented children were more likely to be excluded from the analyses requiring two or more BMI records due to missing data. Further, missing BMI information may not be random and there could be unmeasured reasons for the missingness such as provider, health care setting or child-level factors. Nonetheless, our findings provide valuable information about the potential use of linked Medicaid claims and EHR data to incorporate clinical outcomes into quality measurement compared to traditional medical record review processes with small sample sizes and uncertain generalizability.
      In conclusion, our results demonstrate the importance of addressing gaps in available clinical information by carefully assessing underlying causes including failure to provide and/or document the care. Moreover, our findings indicate that missing clinical information disproportionately affects different subgroups of children, potentially contributing to inequities in assessing the quality of care. Our use of the BMI information demonstrated an increase in children's BMI 1 year after SGAP initiation while controlling for normal growth. The increase in BMI was higher in children who started with a higher BMI. BMI also differed with race-ethnicity and social vulnerability index, with the differences independent of SGAP medication use. Our findings point to opportunities to better capture clinical information, such as BMI, to advance the measurement of health outcomes for children taking SGAP. This information, in turn, can inform the development of quality improvement initiatives to prevent or reduce weight gain among vulnerable children.

      Acknowledgments

      Financial statement: This study was supported by grant U18HS025298 from the Agency for Healthcare Research and Quality . Research reported in this publication also was supported in part by the University of Florida Clinical and Translational Science Institute , which is supported in part by the NIH National Center for Advancing Translational Sciences under award number UL1TR001427 . Dr Sarah Scholle was supported by grant U18HS025296 from the Agency for Healthcare Research and Quality. At the time the paper was written, Dr Michelle Cardel was supported by the National Institute of Health (NIH) National Heart, Lung, and Blood Institute (K01HL141535). Dr Dominick Lemas was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (K01DK115632), and Dr Keith Muller was supported by NIH National Institute of General Medical Sciences grant R01GM121081. The content is solely the responsibility of the authors and does not necessarily represent the official views of the University of Florida's Clinical and Translational Science Institute or the National Institutes of Health.
      The views expressed in this article are those of the authors, and no official endorsement by the Agency for Healthcare Research and Quality (AHRQ), the Centers for Medicare and Medicaid Services (CMS), or the Department of Health and Human Services (DHHS) is intended or should be inferred.
      This article is published as part of a supplement sponsored by the US Department of Health and Human Services, the Centers for Medicare and Medicaid Services, and the Agency for Healthcare Research and Quality.

      References

        • Crystal S
        • Mackie T
        • Fenton MC
        • et al.
        Rapid growth of antipsychotic prescriptions for children who are publicly insured has ceased, but concerns remain.
        Health Aff. 2016; 35: 974-982https://doi.org/10.1377/hlthaff.2016.0064
        • Keast SL
        • Tidmore LM
        • Shropshire D
        • et al.
        Characterization of chronic multiclass psychotropic polypharmacy and psychotherapy in foster care youth in a state Medicaid population.
        J Manag Care Spec Pharm. 2019; 25: 1340-1348https://doi.org/10.18553/jmcp.2019.25.12.1340
        • Leckman-Westin E
        • Finnerty M
        • Scholle SH
        • et al.
        Differences in Medicaid antipsychotic medication measures among children with SSI, foster care, and income-based aid.
        J Manag Care Spec Pharm. 2018; 24: 238-246https://doi.org/10.18553/jmcp.2018.24.3.238
        • Mistry KB
        • Chesley F
        • LLanos K
        • et al.
        Advancing children's health care and outcomes through the pediatric quality measures program.
        Acad Pediatr. 2014; 14: S19-S26https://doi.org/10.1016/j.acap.2014.06.025
      1. All PQMP Measures. Available at: http://www.ahrq.gov/pqmp/measures/all-pqmp-measures.html. Accessed April 14, 2021.

      2. National Collaborative for Innovation in Quality Measurement. NCQA. Available at: https://www.ncqa.org/hedis/reports-and-research/national-collaborative-for-innovation-in-quality-measurement/. Accessed April 27, 2021.

        • Dayabandara M
        • Hanwella R
        • Ratnatunga S
        • et al.
        Antipsychotic-associated weight gain: management strategies and impact on treatment adherence.
        Neuropsychiatr Dis Treat. 2017; 13: 2231-2241https://doi.org/10.2147/NDT.S113099
        • Correll CU
        • Carlson HE.
        Endocrine and metabolic adverse effects of psychotropic medications in children and adolescents.
        J Am Acad Child Adolesc Psychiatry. 2006; 45: 771-791https://doi.org/10.1097/01.chi.0000220851.94392.30
        • Alia KA
        • Wilson DK
        • McDaniel T
        • et al.
        Development of an innovative process evaluation approach for the Families Improving Together (FIT) for weight loss trial in African American adolescents.
        Eval Program Plann. 2015; 49: 106-116https://doi.org/10.1016/j.evalprogplan.2014.12.020
        • Janicke DM
        • Lim CS
        • Mathews AE
        • et al.
        The Community-based Healthy-lifestyle Intervention for Rural Preschools (CHIRP) study: design and methods.
        Contemp Clin Trials. 2013; 34: 187-195https://doi.org/10.1016/j.cct.2012.11.004
        • Janicke DM
        • Sallinen BJ
        • Perri MG
        • et al.
        Comparison of program costs for parent-only and family-based interventions for pediatric obesity in medically underserved rural settings.
        J Rural Health. 2009; 25: 326-330https://doi.org/10.1111/j.1748-0361.2009.00238.x
      3. Meaningful Measures 2.0: Moving from Measure Reduction to Modernization | CMS. Available at:https://www.cms.gov/meaningful-measures-20-moving-measure-reduction-modernization. Accessed October 31, 2021.

      4. Hamlin B, O'Kane M, Barr M, et al. COVID-19 Underscores The Need For Digital Quality Measurement | Health Affairs Blog. Available at: https://www.healthaffairs.org/do/10.1377/hblog20200824.987845/full/. Accessed October 31, 2021.

      5. Medicare-Medicaid Plan (MMP) Chronic Care Improvement Programs & Quality Improvement Projects | CMS. Available at:https://www.cms.gov/Medicare-Medicaid-Coordination/Medicare-and-Medicaid-Coordination/Medicare-Medicaid-Coordination-Office/FinancialAlignmentInitiative/MMPInformationandGuidance/MMPChronicCareImprovementProgramsandQualityImprovementProjects. Accessed November 2, 2021.

        • Hogan WR
        • Shenkman EA
        • Robinson T
        • et al.
        The OneFlorida Data Trust: a centralized, translational research data infrastructure of statewide scope.
        J Am Med Inform Assoc. 2021; (In press)https://doi.org/10.1093/jamia/ocab221
      6. Patient Centered Outcomes Research Institute. PCORnet Common Data Model. Available at:https://pcornet.org/wp-content/uploads/2019/09/PCORnet-Common-Data-Model-v51-2019_09_12.pdf. Accessed September 20, 2020.

      7. 3M. 3M™ Clinical Risk Groups (CRG) Classification System Methodology Overview. Available at: https://apps.3mhis.com/docs/Groupers/Clinical_Risk_Grouping_CRG/methodology_overview/grp401_crg_v2.2_meth_overview.pdf. Accessed December 14, 2021.

      8. U.S. Food and Drug Administration. National Drug Code Directory. Available at: https://www.accessdata.fda.gov/scripts/cder/ndc/index.cfm. Accessed April 22, 2021.

        • Freedman DS
        • Berenson GS.
        Tracking of BMI z scores for severe obesity.
        Pediatrics. 2017; 140e20171072https://doi.org/10.1542/peds.2017-1072
        • Freedman DS
        • Butte NF
        • Taveras EM
        • et al.
        BMI z-scores are a poor indicator of adiposity among 2- to 19-year-olds with very high BMIs, NHANES 1999-2000 to 2013-2014.
        Obesity (Silver Spring). 2017; 25: 739-746https://doi.org/10.1002/oby.21782
      9. SAS Program (ages 0 to < 20 years) | Resources | Growth Chart Training | Nutrition | DNPAO | CDC. 2019. Available at:https://www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas.htm. Accessed March 12, 2021.

      10. USDA ERS - Rural-Urban Continuum Codes. Available at: https://www.ers.usda.gov/data-products/rural-urban-continuum-codes.aspx. Accessed March 12, 2021.

      11. Centers for Disease Control. Social Vulnerability Index. Available at: https://svi.cdc.gov/factsheet.html. Accessed August 27, 2020.

        • Lanigan J.
        Prevention of overweight and obesity in early life.
        Proc Nutr Soc. 2018; 77: 247-256https://doi.org/10.1017/S0029665118000411
        • Romero-Pérez EM
        • González-Bernal JJ
        • Soto-Cámara R
        • et al.
        Influence of a physical exercise program in the anxiety and depression in children with obesity.
        Int J Environ Res Public Health. 2020; 17: 4655https://doi.org/10.3390/ijerph17134655
        • McIntyre RS
        • Mansur RB
        • Lee Y
        • et al.
        Adverse effects of obesity on cognitive functions in individuals at ultra high risk for bipolar disorder: results from the global mood and brain science initiative.
        Bipolar Disord. 2017; 19: 128-134https://doi.org/10.1111/bdi.12491
        • Burstin H
        • Leatherman S
        • Goldmann D.
        The evolution of healthcare quality measurement in the United States.
        J Intern Med. 2016; 279: 154-159https://doi.org/10.1111/joim.12471
        • Ben Amor L.
        Antipsychotics in pediatric and adolescent patients: a review of comparative safety data.
        J Affect Disord. 2012; 138: S22-S30https://doi.org/10.1016/j.jad.2012.02.030