You are currently viewing a new version of our website. To view the old version click .
MDPIMDPI
Search all articles
Healthcare
Download PDF
  • Article
  • Open Access

21 November 2025

Modeling Mental Health Case-Mix for Quality Improvement—A Comparison of Statistical and AI Models

,
,
and
1
Office of Productivity, Efficiency and Staffing, Office of Analytics and Performance Integration, Office of Quality and Patient Safety, Department of Veterans Affairs, Washington, DC 20420, USA
2
Office of Analytics and Performance Integration, Office of Quality and Patient Safety, Department of Veterans Affairs, Washington, DC 20420, USA
3
Department of Computer Science, Siena University, Loudonville, NY 12211, USA
4
Albany Stratton VA Medical Center, Department of Veterans Affairs, Albany, NY 12208, USA
Healthcare2025, 13(23), 3012;https://doi.org/10.3390/healthcare13233012
This article belongs to the Special Issue Applications of Data Mining in Patient Care
Version Notes
Order Reprints
Review Reports

Abstract

Background/Objectives: With the rising prevalence of mental health (MH) disorders, improving the effectiveness and quality of MH care has become increasingly imperative. To improve patient care outcomes, it is essential to accurately assess staffing needs and compare outcomes across providers to identify best practices. However, without a robust case-mix adjustment system that accounts for disease severity, efforts to measure staffing requirements and evaluate patient outcomes are of limited value. This study aimed to develop such a system by leveraging a large study population, more clinically homogeneous groups, and advanced modeling techniques. Methods: In this retrospective population-based study, over two million MH patients (n = 2,088,174) were grouped into 162 clinically homogeneous categories using Clinical Classifications Software Refined (CCSR) to enhance predictive accuracy. We evaluated the performance of four statistical models and four artificial intelligence (AI) models to identify the model that delivered the highest predictive power. Results: Among the statistical models, the Box–Cox regression yielded the highest predictive power (R2 = 0.42; percent of variation explained [PVE] = 0.300). Among the AI models, CatBoost performed best (R2 = 0.458; PVE = 0.311). While the AI models outperformed traditional statistical models, the improvements were modest. Sensitivity analyses confirmed the robustness of these models. Conclusions: Both the Box–Cox and CatBoost models demonstrated superior predictive performance compared to those reported in the literature. These findings suggest that a case-mix system based on either model can be used for risk adjustment to optimize staffing levels and benchmark patient outcomes for quality improvement.

1. Introduction

Mental health (MH) is a fundamental pillar of overall health, deeply intertwined with physical health and overall well-being. Yet, the prevalence of MH disorders continues to rise []. According to the U.S. Substance Abuse and Mental Health Services Administration (SAMHSA) [], in 2023, 23% of adults in the U.S. experienced a mental disorder, while one in 20 experienced a serious mental illness.
Given the increasing prevalence and complexity of mental health conditions, delivering effective and high-quality care is more important than ever. To improve the effectiveness and quality of care, staffing levels need to be optimized, and patient outcomes should be compared across providers to identify best practices [,,,]. To assess staffing needs and patient outcomes, a case-mix system that adjusts for disease severity plays an indispensable role.
Although there is extensive literature on case-mix models for overall disease burden and ambulatory care [,,,], risk-adjustment algorithms specifically tailored to MH remain limited and generally demonstrate low predictive power. For example, in retrospective (concurrent) analyses of actual MH care costs, the highest R-squared reported in the literature was only 0.112 [,].
One major challenge in developing a reliable MH case-mix system is the low diagnostic sensitivity and specificity inherent in MH care. Unlike physical conditions such as diabetes or hypertension, which have clear biomarkers like A1C or blood pressure, mental health diagnoses rely on the DSM (Diagnostic and Statistical Manual of Mental Disorders), introducing some degree of diagnostic ambiguity. This lack of clarity contributes to significant variation in treatment options and associated costs, even among patients with the same diagnosis.
Motivated by these challenges and the pressing need to improve patient care, this study aims to develop a robust mental health case-mix system by leveraging: (1) a large study population, (2) actual patient care costs as the outcome, (3) the enhanced specificity of ICD-10-CM codes and the Clinical Classifications Software Refined (CCSR) framework [], and (4) a range of statistical and AI models.
To enhance the accuracy of the case-mix system, we expanded the CCSR classification into more homogeneous categories and compared the performance of four statistical models and four AI models to evaluate their predictive power. The resulting case-mix system is intended to support health systems in optimizing staffing levels and comparing patient outcomes across providers and hospitals to drive quality improvement.

2. Data and Methods

2.1. Study Population and Data Source

The study population consists of patients with mental health conditions who received care in fiscal year (FY) 2024 in the Veterans Health Administration (VHA). The VHA is the largest integrated health care system in the U.S., providing care to over 9.1 million Veterans enrolled in the VHA health care system at 1380 health care facilities, including 170 Medical Centers and 1193 outpatient clinics []. The VHA health system also maintains a centralized data repository called the Corporate Data Warehouse (CDW), which captures detailed patient demographic, clinical, and cost data.
In this study, we used CCSR developed by AHRQ (Agency for Healthcare Research and Quality) to identify patients with mental health conditions []. CCSR classifies patients into 530 clinically homogeneous groups (e.g., hypertension, diabetes, and COPD) based on International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10 CM) codes. Among the 530 groups, 24 of them are for mental health, which include 889 ICD10 codes. After excluding 109 ICD10 codes such as alcohol-related fatty liver disease that are not directly related to mental health care but included in the 24 CCSR categories, we used 780 ICD-10 codes to identify all patients with MH conditions.
Our primary data source was the VHA CDW. We also incorporated clinical and financial data from MH services provided by non-VA providers that were paid by the VHA, as captured by the Integrated Veteran Care Consolidated Data Sets (IVC CDS).

2.2. Study Variables and Patient Classification

The outcome or dependent variable was the total mental health care cost at the patient level, serving as a proxy for disease severity or burden. This total cost includes the costs of services provided within the VHA system as well as those provided by non-VHA providers that were reimbursed by the VHA.
The independent variables include age, age groups (for nonlinear effects), sex, and 162 clinical groups, which were derived by expanding the 24 CCSR groups to enhance predictive power. The expansion methodology is described in detail elsewhere [,]. Briefly, to increase homogeneity within each clinical group, if a CCSR category included more than 6000 patients (approximately 0.3% of the study population), it was subdivided based on the first digit of the ICD-10 codes. If any resulting subgroup still exceeded 6000 patients, it was further split by the second digit, and so on, until the number of patients would fall below 6000 or the fourth digit was reached. For example, using this process, MBD002 (depressive disorders), which includes 1,544,484 patients, was subdivided into 19 categories.
Although some socioeconomic variables (e.g., private health insurance coverage, marital status, disability ratings) were available, we intentionally excluded them from this study. This is because most existing studies in the literature and commercial case-mix software such as DxCG only use age, sex, and diagnoses as input variables. Using the same input data allows for meaningful comparisons of the predictive performance of different models. Nonetheless, socioeconomic factors can significantly influence health status and should be considered alongside case-mix measures when analyzing staffing levels or comparing patient outcomes.

2.3. Statistical Analyses and Artificial Intelligence (AI) Modeling

Multivariate regression is the most widely used analytical tool for developing case-mix or risk adjustment systems [,,]. After fitting the models, the resulting predicted or expected outcome (cost) serves as a metric for disease burden or severity.
Given that the performance of statistical models can vary significantly depending on the data structure, this study evaluated four commonly used models: ordinary least squares (OLS), Gamma regression with a log link, log-linear regression, and Box–Cox regression. Model performance was assessed using R-squared, percent of variation explained (PVE), mean absolute percentage error (MAPE), and mean absolute error (MAE) [,,,]. These goodness-of-fit metrics were calculated on both the transformed and raw dollar scales. For log-linear and Box–Cox models, the predicted raw dollar costs were obtained using Duan’s nonparametric smearing retransformation to reduce bias [,,].
To guard against overfitting of the models, we randomly split the study population into a development sample (50%) and a validation sample (50%) [,]. The models were fitted on the development sample, and the resulting coefficients were applied to the validation sample to generate predicted values. In addition, we conducted extensive sensitivity analyses, including imposing hierarchies on the relevant expanded CCSR categories [], and testing different development/validation splits (60/40 and 80/20).
In addition to the four statistical models, we evaluated four AI models: Random Forest, LightGBM, XGBoost, and CatBoost. Each AI model has distinct strengths and weaknesses. For example, Random Forest is less prone to overfitting but can struggle with high-cardinality categorical variables and may not capture complex interactions as effectively as boosting methods. LightGBM is highly efficient and fast, especially with large datasets, but it can be sensitive to overfitting. XGBoost delivers strong performance in modeling structured data, but it is slower than LightGBM. CatBoost excels with categorical features and requires less preprocessing, making it particularly user-friendly for health care data, but it may be slower to train.
Although AI has been used in health care to predict future costs or patient outcomes [,,], we are not aware of its application in developing concurrent case-mix models. We evaluated the performance of the four AI models by applying the same metrics used for the statistical models.
All statistical modeling was carried out using SAS Enterprise Guide 8.3, and all the AI models were implemented using Python 3.13.

3. Results

The present study included all 2,088,174 patients who had at least one clinical encounter with MH as the principal diagnosis in FY2024 (either with VHA providers or community providers reimbursed by the VHA). As shown in Table 1, the average patient age was 54.1 years, and 18% were female. The average cost per patient was $7135. The cost difference across age groups was substantial (p < 0.0001), but the difference was relatively small ($255) between male and female patients, although still statistically significant (p < 0.0001).
Table 1. Cost Distribution by Age and Sex.
Table 2 presents the distribution of all MH patients across the 24 CCSR categories. Using these categories and ICD-10-CM codes, we further classified the study population into 162 more clinically homogeneous groups to enhance predictive power. As shown in Table 2, the largest number of patients fell under BMD007 (trauma- and stressor-related disorders), while BMD023 (inhalant-related disorders) had the fewest. Note that patients may have comorbidities across multiple CCSR categories.
Table 2. Mental Health Patients by CCSR Category (FY2024).
In statistical modeling, the development and validation samples (50/50 percent split) produced virtually identical model fit statistics (with R2 values differing only at the fourth decimal place), indicating no overfitting. Thus, all model fit statistics reported in this study are based on the validation sample. Given the large sample size (1,044,087 in the development sample), the forward, backward, and stepwise selection procedures yielded virtually identical results. Therefore, only the results based on the stepwise procedure are reported here. For log-linear and Box–Cox models, we also transformed the predicted values back to the raw dollar scale using Duan’s nonparametric smearing estimator and recalculated the model fit statistics.
Although R2 is widely used to gauge model fit, it greatly overestimates the percentage of variation explained by the model []. Therefore, we also report PVE (Percent of Variation Explained), which is more informative for model selection. As shown in Table 3, OLS yielded an R2 of 0.407, however, only 19.3% of the variation in cost was explained by the independent variables.
Table 3. Predictive Power of the Statistical Models.
The Gamma regression with a log link, designed for positively skewed data, was outperformed by three other models. Log-linear regression had reasonable predictive power, but it performed poorly after the predicted values were transformed back to the raw dollar scale. These findings were consistent with other studies [,]. Among the four models, Box–Cox regression offered the highest predictive power: R2 = 0.458 and PVE = 0.300 on the raw scale.
The model fit statistics of the four AI models are reported in Table 4. As shown, the predictive power of the four models was comparable: Random Forest yielded the lowest predictive power (R2 = 0.428, PVE = 0.291), and CatBoost produced the highest (R2 = 0.458, PVE = 0.311). Overall, AI models outperformed statistical models in terms of predictive power, but the improvement was small, which is consistent with findings from other studies [,,].
Table 4. Predictive Power of the AI Models.
We conducted extensive sensitivity analyses to ensure the robustness of the models. For example, we imposed clinical hierarchies on the expanded CCSR groups, but this had minimal impact on model fit metrics. We also subdivided the 162 expanded CCSR groups to the fifth digit of the ICD-10-CM codes; however, the improvement in predictive power was negligible.
For the Box–Cox regression, we tested multiple values of the transformation parameter (λ) and confirmed that λ = 0.548 yielded the highest predictive performance, as reported in Table 3. Among the AI models, we varied numerical hyperparameters by ±10% to identify the optimal values that maximized predictive power on the validation samples. For instance, in the CatBoost model, after optimization, the final hyperparameters were set as: iterations = 1000, learning_rate = 0.1, depth = 6, loss_function = ‘RMSE’, and verbose = 100.

4. Discussion

Mental health disorders are increasingly prevalent worldwide and are responsible for immense suffering, reduced quality of life, adverse physical health, increased mortality, and staggering economic and social costs [,,]. Given the rising MH care needs and limited resources, optimizing MH staffing levels and comparing patient care outcomes across providers to identify best practices have become imperative. However, assessing staffing needs and analyzing outcomes can be counterproductive without a robust case-mix system that accounts for patient disease severity.
By leveraging a large sample size, more granular clinical groups, and best-fit statistical and AI models, this study achieved a predictive power that is four times higher than what has been reported in the literature (R2 ≤ 0.112) [,]. Our extensive sensitivity analyses demonstrated that the two best performing models (Box–Cox and CatBoost) were robust and generalizable. The significant improvement in predictive power has practical implications for health systems seeking to optimize staffing, allocate resources, and benchmark outcomes in mental health care. By providing a more accurate adjustment for patient complexity, this case-mix system can support fairer comparisons across providers and inform value-based care initiatives.
Notably, although AI models outperformed statistical models in this study, the improvement in predictive power was modest, consistent with findings from previous research [,,]. One possible explanation is that the relationship between the outcome and predictors may not be highly nonlinear or complex, which are conditions under which AI models typically excel. Moreover, the input data (age, sex, and diagnoses) are well-defined and structured, limiting the advantages of AI models that are designed to process complex or unstructured data. Future research could explore hybrid modeling approaches that integrate traditional statistical methods with AI techniques to further improve predictive performance.
Despite the improved predictive power, the study has limitations. The findings are based on data from the VHA, where the majority of patients in the study were male (82%), which could raise concerns about the generalizability of the models to other populations or care settings. However, after excluding sex from the models in our sensitivity analyses, the decrease in R2 across different models was less than 0.01. Therefore, the impact on generalizability is likely minimal. In addition, some patients in this study may have used services paid for by Medicare, but the clinical and financial data for those services were not included in this study due to the time lag of data availability. Nonetheless, it is reasonable to assume that the predictive power of the models would increase if Medicare data were available and included.

5. Conclusions

In summary, this study presents a scalable, data-driven approach to mental health case-mix modeling that addresses a critical need in the field. By integrating refined diagnostic groupings with advanced statistical and AI techniques, the resulting case-mix system offers a practical and robust method to account for disease severity when assessing staffing requirements and benchmarking patient care outcomes for quality improvement.

Author Contributions

Conceptualization, J.G. and T.L.B.; Methodology, J.G., T.L. and S.L.F.; Software, J.G. and T.L.; Validation, J.G., T.L.B., T.L. and S.L.F.; Formal analysis, J.G. and T.L.; Investigation, J.G., T.L.B., T.L. and S.L.F.; Resources, T.L.B.; Data curation, J.G. and T.L.B.; Writing—original draft, J.G.; Writing—review and editing, J.G., T.L.B., T.L. and S.L.F.; Supervision, J.G. and T.L.B.; Project administration, T.L.B. and S.L.F. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

The study was approved for publication by the Albany Stratton VA Medical Center R&D Office and utilized secondary, de-identified patient care data obtained from the Veterans Health Administration’s Corporate Data Warehouse (CDW) and the Integrated Veteran Care Consolidated Data Sets (IVC CDS). As the analysis involved only existing patient records without direct contact or intervention, Institutional Review Board (IRB) approval was not required in accordance with VA Title 38, Section 16.101(b)(4).

Data Availability Statement

The data presented in this study are available on request from the corresponding author due to the scale and sensitive nature of the data set.

Acknowledgments

This material is based upon work supported (or supported in part) by the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Dykxhoorn, J.; Solmi, F.; Walters, K.; Gnani, S.; Lazzarino, A.; Kidger, J.; Kirkbride, J.B.; Osborn, D.P.J. Common mental disorders in young adults: Temporal trends in primary care episodes and self-reported symptoms. BMJ Ment. Health 2025, 28, e301457. [Google Scholar] [CrossRef]
  2. Substance Abuse and Mental Health Services Administration. Key Substance Use and Mental Health Indicators in the United States: Results from the 2023 National Survey on Drug Use and Health; HHS Publication No. PEP24-07-021, NSDUH Series H-59; Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration: Rockville, MD, USA, 2024. Available online: https://www.samhsa.gov/data/sites/default/files/reports/rpt47095/National%20Report/National%20Report/2023-nsduh-annual-national.pdf (accessed on 5 November 2025).
  3. Boden, M.; Smith, C.A.; Trafton, J.A. Investigation of population-based mental health staffing and efficiency-based mental health productivity using an information-theoretic approach. PLoS ONE 2021, 16, e0256268. [Google Scholar] [CrossRef] [PubMed] [PubMed Central]
  4. Smith, C.A.; Boden, M.T.; Trafton, J.A. Outpatient provider staffing ratios: Binary recursive models associated with quality, access, and satisfaction. Psychol. Serv. 2023, 20, 137–143. [Google Scholar] [CrossRef] [PubMed]
  5. Aragón, M.J.; Gravelle, H.; Castelli, A.; Goddard, M.; Gutacker, N.; Mason, A.; Rowen, D.; Mannion, R.; Jacobs, R. Measuring the overall performance of mental healthcare providers. Soc. Sci. Med. 2024, 344, 116582. [Google Scholar] [CrossRef] [PubMed]
  6. Kilbourne, A.M.; Beck, K.; Spaeth-Rublee, B.; Ramanuj, P.; O’Brien, R.W.; Tomoyasu, N.; Pincus, H.A. Measuring and improving the quality of mental health care: A global perspective. World Psychiatry 2018, 17, 30–38. [Google Scholar] [CrossRef] [PubMed] [PubMed Central]
  7. Schone, E.; Brown, R.S. Risk Adjustment: What Is the Current State of the Art and How Can It Be Improved? Robert Wood Johnson Foundation: Princeton, NJ, USA, 2013. Available online: https://www.physicianprofiling.ch/ProfilingStateAndImprove2013Overview.pdf (accessed on 5 November 2025).
  8. Thomas, J.W.; Grazier, K.L.; Ward, K. Comparing accuracy of risk-adjustment methodologies used in economic profiling of physicians. Inquiry 2004, 41, 218–231. [Google Scholar] [CrossRef] [PubMed]
  9. Centers for Medicare & Medicaid Services. CMS Risk Adjustment Discussion Paper. Available online: https://www.cms.gov/CCIIO/Resources/Forms-Reports-and-Other-Resources/Downloads/RA-March-31-White-Paper-032416.pdf (accessed on 5 November 2025).
  10. Gao, J.; Moran, E.; Almenoff, P.L. Case-mix for performance management: A risk algorithm based on ICD-10-CM. Med. Care 2018, 56, 537–543. [Google Scholar] [CrossRef] [PubMed]
  11. Sloan, K.L.; Montez-Rath, M.E.; Spiro, A., 3rd; Christiansen, C.L.; Loveland, S.; Shokeen, P.; Herz, L.; Eisen, S.; Breckenridge, J.N.; Rosen, A.K. Development and validation of a psychiatric case-mix system. Med. Care 2006, 44, 568–580. [Google Scholar] [CrossRef] [PubMed]
  12. Tran, N.; Poss, J.W.; Perlman, C.; Hirdes, J.P. Case-mix classification for mental health care in community settings: A scoping review. Health Serv. Insights 2019, 12, 1178632919862248. [Google Scholar] [CrossRef]
  13. Agency for Healthcare Research and Quality (AHRQ). Clinical Classifications Software Refined (CCSR). 2024. Available online: https://hcup-us.ahrq.gov/toolssoftware/ccsr/dxccsr.jsp (accessed on 5 November 2025).
  14. Veterans Health Administration. About VHA. Available online: https://www.va.gov/health/aboutvha.asp#:~:text=The%20Veterans%20Health%20Administration%20(VHA,the%20VA%20health%20care%20program (accessed on 5 November 2025).
  15. Gao, J.; Moran, E.; Schwartz, A.; Ruser, C. Case-mix for assessing primary care value (CPCV). Health Serv. Manag. Res. 2020, 33, 200–206. [Google Scholar] [CrossRef]
  16. Gao, J.; Moran, E.; Higgins, D.S., Jr.; Mecher, C. Predicting high-risk and high-cost patients for proactive intervention. Med. Care 2022, 60, 610–615. [Google Scholar] [CrossRef]
  17. Iezzoni, L.I. (Ed.) Risk Adjustment for Measuring Health Care Outcomes, 3rd ed.; Health Administration Press: Chicago, IL, USA, 2003. [Google Scholar]
  18. Ellis, R.P. Risk adjustment in health care markets: Concepts and applications. In Financing Health Care; John Wiley & Sons: Hoboken, NJ, USA, 2008; pp. 177–222. [Google Scholar]
  19. Winkelman, R.; Mehmud, S. A Comparative Analysis of Claims-Based Tools for Health Risk Assessment; Society of Actuaries: Schaumburg, IL, USA, 2007. [Google Scholar]
  20. Gao, J. R-squared (R2)—How much variation is explained? Res. Methods Med. Health Sci. 2024, 5, 104–109. [Google Scholar] [CrossRef]
  21. Pope, G.C.; Kautter, J.; Ellis, R.P.; Ash, A.S.; Ayanian, J.Z.; Iezzoni, L.I.; Ingber, M.J.; Levy, J.M.; Robst, J. Risk adjustment of Medicare capitation payments using the CMS-HCC model. Health Care Financ. Rev. 2004, 25, 119–141. [Google Scholar]
  22. Ash, A.S.; Ellis, R.P.; Pope, G.C.; Ayanian, J.Z.; Bates, D.W.; Burstin, H.; Iezzoni, L.I.; MacKay, E.; Yu, W. Using diagnoses to describe populations and predict costs. Health Care Financ. Rev. 2000, 21, 7–28. [Google Scholar]
  23. Duan, N.; Manning, W.G.; Morris, C.N.; Newhouse, J.P. A comparison of alternative models for the demand for medical care. J. Bus. Econ. Stat. 1983, 1, 115–126. [Google Scholar] [CrossRef]
  24. Mihaylova, B.; Briggs, A.; O’Hagan, A.; Thompson, S.G. Review of statistical methods for analysing healthcare resources and costs. Health Econ. 2011, 20, 897–916. [Google Scholar] [CrossRef]
  25. Jones, A. Models for health care. In The Oxford Handbook of Economic Forecasting; Hendry, D., Clements, M., Eds.; Oxford University Press: Oxford, UK, 2010; pp. 625–654. [Google Scholar]
  26. Harrell, F.E.; Lee, K.L.; Mark, D.B. Tutorial in biostatistics multivariable prognostic models: Issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat. Med. 1996, 15, 361–387. [Google Scholar] [CrossRef]
  27. Efron, B. Estimating the error rate of a prediction rule: Improvement on cross-validation. J. Am. Stat. Assoc. 1983, 78, 316–331. [Google Scholar] [CrossRef]
  28. Riis, A.H.; Kristensen, P.K.; Lauritsen, S.M.; Thiesson, B.; Jørgensen, M.J. Using explainable artificial intelligence to predict potentially preventable hospitalizations: A population-based cohort study in Denmark. Med. Care 2023, 61, 226–236. [Google Scholar] [CrossRef] [PubMed]
  29. Dixon, D.; Sattar, H.; Moros, N.; Kesireddy, S.R.; Ahsan, H.; Lakkimsetti, M.; Fatima, M.; Doshi, D.; Sadhu, K.; Junaid Hassan, M. Unveiling the influence of AI predictive analytics on patient outcomes: A comprehensive narrative review. Cureus 2024, 16, e59954. [Google Scholar] [CrossRef] [PubMed]
  30. Nong, P.; Adler-Milstein, J.; Apathy, N.C.; Holmgren, A.J.; Everson, J. Current use and evaluation of artificial intelligence and predictive models in US hospitals. Health Aff. 2025, 44, 90–98. [Google Scholar] [CrossRef]
  31. Gao, J.; Moran, E.; Li, Y.F.; Almenoff, P.L. Predicting potentially avoidable hospitalizations. Med. Care 2014, 52, 164–171. [Google Scholar] [CrossRef]
  32. Desai, R.J.; Wang, S.V.; Vaduganathan, M.; Evers, T.; Schneeweiss, S. Comparison of machine learning methods with traditional models for use of administrative claims with electronic medical records to predict heart failure outcomes. JAMA Netw. Open 2020, 3, e1918962. [Google Scholar] [CrossRef]
  33. Funk, M.; Saraceno, B.; Drew, N.; Faydi, E. Integrating mental health into primary healthcare. Ment. Health Fam. Med. 2008, 5, 5–8. [Google Scholar] [PubMed] [PubMed Central]
  34. Henderson, G. Addressing the public’s mental health. J. Public Health 2015, 37, 370–372. [Google Scholar] [CrossRef] [PubMed]
  35. Moitra, M.; Owens, S.; Hailemariam, M.; Wilson, K.S.; Mensa-Kwao, A.; Gonese, G.; Kamamia, C.K.; White, B.; Young, D.M.; Collins, P.Y. Global Mental Health: Where We Are and Where We Are Going. Curr. Psychiatry Rep. 2023, 25, 301–311. [Google Scholar] [CrossRef] [PubMed]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Journal Statistics
Article metric data becomes available approximately 24 hours after publication online.
Download PDF
Narrative Review on Post-Stroke Outcomes Through Recognition of Frailty, Sarcopenia, and Palliative Care Needs
Previous
Efficacy of Baduanjin Versus Brisk Walking on Cognitive and Physical Functions in Schizophrenia: A Three-Arm Randomized Controlled Trial
Next