Predictors Of Health Care Utilization Research Articles

Machine Learning-Based Prediction for High Health Care Utilizers by Using a Multi-Institutional Diabetes Registry: Model Training and Evaluation.

The cost of health care in many countries is increasing rapidly. There is a growing interest in using machine learning for predicting high health care utilizers for population health initiatives. Previous studies have focused on individuals who contribute to the highest financial burden. However, this group is small and represents a limited opportunity for long-term cost reduction. We developed a collection of models that predict future health care utilization at various thresholds. We utilized data from a multi-institutional diabetes database from the year 2019 to develop binary classification models. These models predict health care utilization in the subsequent year across 6 different outcomes: patients having a length of stay of ≥7, ≥14, and ≥30 days and emergency department attendance of ≥3, ≥5, and ≥10 visits. To address class imbalance, random and synthetic minority oversampling techniques were employed. The models were then applied to unseen data from 2020 and 2021 to predict health care utilization in the following year. A portfolio of performance metrics, with priority on area under the receiver operating characteristic curve, sensitivity, and positive predictive value, was used for comparison. Explainability analyses were conducted on the best performing models. When trained with random oversampling, 4 models, that is, logistic regression, multivariate adaptive regression splines, boosted trees, and multilayer perceptron consistently achieved high area under the receiver operating characteristic curve (>0.80) and sensitivity (>0.60) across training-validation and test data sets. Correcting for class imbalance proved critical for model performance. Important predictors for all outcomes included age, number of emergency department visits in the present year, chronic kidney disease stage, inpatient bed days in the present year, and mean hemoglobin A1c levels. Explainability analyses using partial dependence plots demonstrated that for the best performing models, the learned patterns were consistent with real-world knowledge, thereby supporting the validity of the models. We successfully developed machine learning models capable of predicting high service level utilization with strong performance and valid explainability. These models can be integrated into wider diabetes-related population health initiatives.

Social Determinants of Health and Healthcare Utilization Disparities among Older Adults with and Without Cognitive Impairment.

The purpose of this study was to determine the healthcare utilization patterns in a national sample of older adults across several social determinants of health factors (ethnicity, gender, race, education) with normal and dementia/impaired cognition. We used datasets from the Health and Retirement Study (HRS, 2018) to evaluate healthcare utilization, including metrics such as hospital and nursing home stays, hospice care, and number of visits to the doctor. Logistic models were used to predict healthcare utilization separately in those with normal cognition and dementia. Our final sample comprised 15,607 adults (mean age: 65.2 normal cognition, mean age 71.5 dementia). Hispanics with normal cognition were less likely to stay in a hospital than non-Hispanic respondents (OR: 0.52-0.71, p<0.01). Being female was associated with a higher risk for shorter nursing home days (OR: 1.41, p<0.01) and doctor visits (OR: 1.63-2, p<0.01) in cognitively normal older adults. Being female was associated with a lower risk for hospital stay in those with dementia (OR: 0.50-0.78, p<0.01). Respondents identifying as Black or other races with dementia were less likely to experience nursing home days (OR: 0.42, p<0.04). Black respondents with normal cognition were less likely to experience doctor visits (OR: 0.32-0.37, p<0.01). Those with more than a high school education in both groups were more likely to experience doctors' visits. The study points to the continued disparities in healthcare utilization linked to participants' social determinants of health factors and cognition.

The Relationship Between Hyperalbuminemia and Unscheduled Medical Visits: A Retrospective Cohort Study.

Introduction Hyperalbuminemia, defined as elevated serum albumin levels, may influence healthcare utilization, particularly unscheduled medical visits. The sympathetic nervous system (SNS) regulates serum albumin, which is crucial for maintaining oncotic pressure and substance transport. SNS instability, linked to chronic diseases, can impact albumin levels. This study investigates the association between hyperalbuminemia and unscheduled medical visits in community hospital outpatient departments, aiming to establish its potential as a predictor of healthcare utilization. Methods This retrospective cohort study utilized electronic medical records from Unnan City Hospital, Japan, from September 2021 to August 2023. Participants were over 15 years old and had albumin data available, excluding those with acute albumin conditions. The case group consisted of 321 hyperalbuminemia patients (serum albumin ≥ 5 g/dL), matched monthly with 16 controls. Data on demographics, chronic diseases, and unscheduled medical visits were collected. Multivariate logistic regression analyzed the association between hyperalbuminemia and unscheduled medical visits. Results Among 716 participants, the hyperalbuminemia group (mean age 59.13 years) was younger than the non-hyperalbuminemia group (mean age 74.36 years). Hyperalbuminemia patients had a higher BMI, pulse rate, and prevalence of diabetes, dyslipidemia, and brain stroke. Significant associations were found between hyperalbuminemia and unscheduled medical visits (OR 2.35, 95% CI 1.56-3.53, p < 0.001), age, BMI, pulse rate, and brain stroke. Conclusion Hyperalbuminemia is significantly associated with increased unscheduled medical visits in rural outpatient settings. Routine serum albumin assessments can aid in risk stratification and personalized care, potentially reducing acute healthcare needs. Future research should explore underlying mechanisms and broader populations to enhance clinical applications.

American clusters: using machine learning to understand health and health care disparities in the United States.

Health and health care access in the United States are plagued by high inequality. While machine learning (ML) is increasingly used in clinical settings to inform health care delivery decisions and predict health care utilization, using ML as a research tool to understand health care disparities in the United States and how these are connected to health outcomes, access to health care, and health system organization is less common. We utilized over 650 variables from 24 different databases aggregated by the Agency for Healthcare Research and Quality in their Social Determinants of Health (SDOH) database. We used k-means-a non-hierarchical ML clustering method-to cluster county-level data. Principal factor analysis created county-level index values for each SDOH domain and 2 health care domains: health care infrastructure and health care access. Logistic regression classification was used to identify the primary drivers of cluster classification. The most efficient cluster classification consists of 3 distinct clusters in the United States; the cluster having the highest life expectancy comprised only 10% of counties. The most efficient ML clusters do not identify the clusters with the widest health care disparities. ML clustering, using county-level data, shows that health care infrastructure and access are the primary drivers of cluster composition.

The Prediction of Healthcare Utilization by Frailty and Disability among Dutch Community-Dwelling People Aged 75 Years or Older.

Population aging is occurring worldwide. As a result, frailty and disability are in the full interest of practice, policy, and science. An increase in healthcare utilization is an adverse outcome of frailty and disability. The aim of the present study was the prediction of six indicators of healthcare utilization by frailty and disability measures. The six indicators of healthcare utilization of interest were: use of informal care, number of visits to a general practitioner, hospital admission, receiving nursing care, receiving personal care, and contacts with (health)care professionals. We used a sample of 484 people that was randomly drawn from the municipality of Roosendaal (the Netherlands), a municipality with 78,000 inhabitants. A subset of people who completed the Tilburg Frailty Indicator (TFI) at baseline and the Groningen Activity Restriction Scale (GARS) questionnaires was used with a nine-year followup. We used generalized estimation equations (GEE) to predict the six indicators with the frailty measures (physical, psychological, and social scores) and disability measures (ADL and IADL scores). We also performed GEE analyses adjusted for age, gender, and multimorbidity from part A of the TFI at baseline. The significant predictors were different for each indicator. However, the physical frailty score, the ADL score, and the IADL score often emerged as significant predictors. These three variables predicted several indicators of healthcare utilization: use of informal care, number of visits to a general practitioner, hospital admission, receiving nursing care, receiving personal care, and contacts with (health)care professionals. The social score was found to be significant for the indicator use of disciplines. In conclusion, our study showed that in particular physical frailty, and ADL and IADL disability predicted healthcare utilization in community-dwelling people aged 75 years or older. It is important that care and welfare professionals pay attention to physical frailty and both ADL and IADL disability aiming to diminish frailty and disability and preventing intensive healthcare utilization and related costs. Future research will have to focus on more representative Dutch municipalities in order to get a more reliable and accurate picture of the disciplines used by people with frailty and disability.

The liver frailty index is a predictor of healthcare utilization after liver transplantation in older adults.

Older adults have higher healthcare utilization after liver transplantation (LT), yet objective risk stratification tools in this population are lacking. We evaluated the Liver Frailty Index (LFI) as one potential tool. Ambulatory LT candidates ≥65 years without hepatocellular carcinoma (HCC) who underwent LT from 1/2012 to 6/2022 at 8 U.S. centers were included. Estimates of the difference in median using quantile regression were used to assess the adjusted association between LFI and hospitalized days within 90 days post-LT. Of 131 LT recipients, median (interquartile range [IQR]) (1st -3rd quartiles) age was 68 years (66-70); median pre-LT MELD-Na was 19 (15-24). Median LFI was 4.1 (3.6-4.7); 27% were frail (LFI≥4.5). Median hospitalized days within 90 days post-LT was 11 (7-20). Compared with non-frail patients, frail patients were hospitalized for a median of 5 days longer post-LT (95% CI .30-9.7, p=.04). Each .5 unit increase in pre-LT LFI was associated with an increase of 1.16 days (95%CI .42-2.69, p=.02) in hospitalized days post-LT. Among older adults undergoing LT, frailty was associated with more hospitalized days within 90 days after LT. The LFI can identify older adults who might benefit from pre-LT or early post-LT programs which may reduce post-LT healthcare utilization, such as early rehabilitation or post-hospital discharge programs.

Predictors of Healthcare Utilization in Family Caregivers of Persons With a Primary Malignant Brain Tumor.

BACKGROUND: Negative physical health results from the emotional stress of providing care to a family member with a primary malignant brain tumor; however, the downstream effects on caregivers' healthcare utilization (HCU) are unknown. This analysis examined associations between caregivers' emotional health and markers of HCU during the 6 months after patients' diagnoses. METHODS: Caregivers' self-report HCU data from a longitudinal study with 116 neuro-oncology caregivers were analyzed. Healthcare utilization was operationalized as number of prescription medications, reporting visits to primary care providers (PCPs), nature of PCP visit, number of comorbid conditions, and change in comorbid conditions. Potential predictors were caregivers' depressive symptoms (Center for Epidemiologic Studies-Depression Scale), hours providing care per day, mastery (Pearlin and Schooler), and burden (Caregiver Reaction Assessment). Logistic mixed effects modeling were used. RESULTS : Caregivers with higher levels of depressive symptoms ( P < .01), anxiety ( P = .02), burden related to schedule ( P = .02), and abandonment ( P < .01) were more likely to report worsening comorbid conditions. Those with higher mastery ( P = .02) were less likely to report worsening comorbid conditions. Caregivers who had a PCP visit and reported higher burden related to feelings of self-esteem ( P = .03) were more likely to report an illness-related visit. CONCLUSION : Findings suggest a relationship between neuro-oncology caregivers' emotional health and their HCU. Data highlight the importance of caregivers' PCPs identifying caregivers at risk for deteriorating health and increased HCU and intervene to ensure caregivers' self-care.

Patterns and Predictors of Health Care Utilization After Pediatric Concussion: A Retrospective Cohort Study

ObjectiveTo characterize types, duration, and intensity of health care utilization following pediatric concussion and to identify risk factors for increased post-concussion utilization. MethodsA retrospective cohort study of children 5 to 17 years old diagnosed with acute concussion at a quaternary center pediatric emergency department or network of associated primary care clinics. Index concussion visits were identified using International Classification of Diseases, Tenth Revision, Clinical Modification codes. We analyzed patterns of health care visits 6 months before and after the index visit using interrupted time-series analyses. The primary outcome was prolonged concussion-related utilization, defined as having ≥1 follow-up visits with a concussion diagnosis more than 28 days after the index visit. We used logistic regressions to identify predictors of prolonged concussion-related utilization. ResultsEight hundred nineteen index visits (median [interquartile range] age, 14 [11–16] years; 395 [48.2%] female) were included. There was a spike in utilization during the first 28 days after the index visit compared to the pre-injury period. Premorbid headache/migraine disorder (adjusted odds ratio (aOR) 2.05, 95% confidence interval [CI] 1.09–3.89) and top quartile pre-injury utilization (aOR 1.90, 95% CI 1.02–3.52) predicted prolonged concussion-related utilization. Premorbid depression/anxiety (aOR 1.55, 95% CI 1.31–1.83) and top quartile pre-injury utilization (aOR 2.29, 95% CI 1.95–2.69) predicted increased utilization intensity. ConclusionsHealth care utilization is increased during the first 28 days after pediatric concussion. Children with premorbid headache/migraine disorders, premorbid depression/anxiety, and high baseline utilization are more likely to have increased post-injury health care utilization. This study will inform patient-centered treatment but may be limited by incomplete capture of post-injury utilization and generalizability.

Quantitative CT Lung Imaging and Machine Learning Improves Prediction of Emergency Room Visits and Hospitalizations in COPD.

Predicting increased risk of future healthcare utilization in chronic obstructive pulmonary disease (COPD) patients is an important goal for improving patient management. Our objective was to determine the importance of computed tomography (CT) lung imaging measurements relative to other demographic and clinical measurements for predicting future health services use with machine learning in COPD. In this retrospective study, lung function measurements and chest CT images were acquired from Canadian Cohort of Obstructive Lung Disease study participants from 2010 to 2017 (https://clinicaltrials.gov, NCT00920348). Up to two follow-up visits (1.5- and 3-year follow-up) were performed and participants were asked for details related to healthcare utilization. Healthcare utilization was defined as any COPD hospitalization or emergency room visit due to respiratory problems in the 12 months prior to the follow-up visits. CT analysis was performed (VIDA Diagnostics Inc.); a total of 108 CT quantitative emphysema, airway and vascular measurements were investigated. A hybrid feature selection method with support vector machine classifier was used to predict healthcare utilization. Performance was determined using accuracy, F1-measure and area under the receiver operating characteristic curve (AUC) and Matthews's correlation coefficient (MC). Of the 527 COPD participants evaluated, 179 (35%) used healthcare services at follow-up. There were no significant differences between the participants with or without healthcare utilization at follow-up for age (p=0.50), sex (p=0.44), BMI (p=0.05) or pack-years (p=0.76). The accuracy for predicting subsequent healthcare utilization was 80% ± 3% (F1-measure=74%, AUC=0.80, MC=0.6) when all measurements were considered, 76% ± 6% (F1-measure=72%, AUC=0.77, MC=0.55) for CT measurements alone and 65% ± 5% (F1-measure=60%, AUC=0.67, MC=0.34) for demographic and lung function measurements alone. The combination of CT lung imaging and conventional measurements leads to greater prediction accuracy of subsequent health services use than conventional measurements alone, and may provide needed prognostic information for patients suffering from COPD.

Differential impact of the COVID-19 pandemic on primary care utilization related to common mental disorders in four European countries: A retrospective observational study.

The COVID-19 pandemic is commonly believed to have increased common mental disorders (CMD, i.e., depression and anxiety), either directly due to COVID-19 contractions (death of near ones or residual conditions), or indirectly by increasing stress, economic uncertainty, and disruptions in daily life resulting from containment measure. Whereas studies reporting on initial changes in self-reported data frequently have reported increases in CMD, pandemic related changes in CMD related to primary care utilization are less well known. Analyzing time series of routinely and continuously sampled primary healthcare data from Sweden, Norway, Netherlands, and Latvia, we aimed to characterize the impact of the pandemic on CMD recorded prevalence in primary care. Furthermore, by relating these changes to country specific time-trajectories of two classes of containment measures, we evaluated the differential impact of containment strategies on CMD rates. Specifically, we wanted to test whether school restrictions would preferentially affect age groups corresponding to those of school children or their parents. For the four investigated countries, we collected time-series of monthly counts of unique CMD patients in primary healthcare from the year 2015 (or 2017) until 2021. Using pre-pandemic timepoints to train seasonal Auto Regressive Integrated Moving Average (ARIMA) models, we predicted healthcare utilization during the pandemic. Discrepancies between observed and expected time series were quantified to infer pandemic related changes. To evaluate the effects of COVID-19 measures on CMD related primary care utilization, the predicted time series were related to country specific time series of levels of social distancing and school restrictions. In all countries except Latvia there was an initial (April 2020) decrease in CMD care prevalence, where largest drops were found in Sweden (Prevalence Ratio, PR = 0.85; 95% CI 0.81-0.90), followed by Netherlands (0.86; 95% CI 0.76-1.02) and Norway (0.90; 95% CI 0.83-0.98). Latvia on the other hand experienced increased rates (1.25; 95% CI 1.08-1.49). Whereas PRs in Norway and Netherlands normalized during the latter half of 2020, PRs stayed low in Sweden and elevated in Latvia. The overall changes in PR during the pandemic year 2020 was significantly changed only for Sweden (0.91; 95% CI 0.90-0.93) and Latvia (1.20; 95% CI 1.14-1.26). Overall, the relationship between containment measures and CMD care prevalence were weak and non-significant. In particular, we could not observe any relationship of school restriction to CMD care prevalence for the age groups best corresponding to school children or their parents. Common mental disorders prevalence in primary care decreased during the initial phase of the COVID-19 pandemic in all countries except from Latvia, but normalized in Norway and Netherlands by the latter half of 2020. The onset of the pandemic and the containment strategies were highly correlated within each country, limiting strong conclusions on whether restriction policy had any effects on mental health. Specifically, we found no evidence of associations between school restrictions and CMD care prevalence. Overall, current results lend no support to the common belief that the pandemic severely impacted the mental health of the general population as indicated by healthcare utilization, apart from in Latvia. However, since healthcare utilization is affected by multiple factors in addition to actual need, future studies should combine complementary types of data to better understand the mental health impacts of the pandemic.

Disparities in healthcare utilization and depression among young mothers: The role of family functioning.

Extant research suggests that disparities in healthcare utilization and experiences of mental health disorders such as depression exist across sociodemographic groups. One understudied pathway that may have significant implications for these disparities is the role of family functioning. How families interact, communicate, and adapt may vary, but these relationships remain integral for vulnerable subsets of the population due to their seemingly protective properties. Yet, few studies have examined the relationship between family functioning and health. The aim of this study was to explore family functioning as a predictor for healthcare utilization and depression among young mothers and as a moderator of disparities for each of these outcomes. This analysis uses data from a prospective cohort study which was conducted with 296 pregnant young women recruited from obstetrics clinics in Connecticut between July 2007 and February 2011 and followed 1-year postpartum. Logistic regression models were used to assess family functioning, race/ethnicity, and immigrant status as predictors of healthcare utilization and depression. Family functioning was evaluated as a moderator for both outcomes. Black and Latina mothers had decreased odds of using physical healthcare services (odds ratioBlack = 0.13, p < 0.001; odds ratioLatina = 0.31, p = 0.02). Family functioning was associated with decreased odds of using mental healthcare services (odds ratio = 0.97, p = 0.04) and had a protective effect against depression (odds ratio = 0.96, p = 0.003). Family functioning significantly moderated the relationship between immigrant status and physical healthcare utilization, Black race/ethnicity and physical healthcare utilization, and Other race/ethnicity and depression. Family functioning is a significant factor associated with both healthcare utilization and depression among young mothers. It should be a strongly considered component within healthcare settings to mitigate risks among vulnerable populations.

Developing a Model to Predict High Health Care Utilization Among Patients in a New York City Safety Net System.

Health care facilities use predictive models to identify patients at risk of high future health care utilization who may benefit from tailored interventions. Previous predictive models that have focused solely on inpatient readmission risk, relied on commercial insurance claims data, or failed to incorporate social determinants of health may not be generalizable to safety net hospital populations. To address these limitations, we developed a payer-agnostic risk model for patients receiving care at the largest US safety net hospital system. We transformed electronic health record and administrative data from 833,969 adult patients who received care during July 2016-July 2017 into demographic, utilization, diagnosis, medication, and social determinant variables (including homelessness and incarceration history) to predict health care utilization during the following year.We selected the final model by developing and validating multiple classification and regression models predicting 10+ acute days, 5+ acute days, or continuous acute days. We compared a portfolio of performance metrics while prioritizing positive predictive value for patients whose predicted utilization was among the top 1% to maximize clinical utility. The final model predicted continuous number of acute days and included 17 variables. For the top 1% of high acute care utilizers, the model had a positive predictive value of 47.6% and sensitivity of 17.3%. Previous health care utilization and psychosocial factors were the strongest predictors of future high acute care utilization. We demonstrated a feasible approach to predictive high acute care utilization in a safety net hospital using electronic health record data while incorporating social risk factors.

Assessment of transition readiness to predict health care utilization during transition to adult care in sickle cell disease

ABSTRACT Background Transition-age patients with sickle cell disease (SCD) are at risk for poor outcomes associated with incomplete transition readiness and neurocognitive deficits. Study objectives were to: 1) test if a SCD-specific measure of self-management skills was associated with transition outcomes and 2) evaluate if caregiver-reported executive functioning was associated with self-management skills and transition outcomes among youth with SCD. Research Design and Methods Youth/caregivers were selected from a longitudinal cohort study. Caregivers completed the Behavior Rating Inventory of Executive Function (BRIEF); caregivers and youth completed the Self-Management Skills Checklist (SMSC) at a median age of 16.8±0.6 years. Non-parametric tests compared SMSC and transition outcomes. Regression assessed the incremental validity of SMSC in predicting transition outcomes. Results In total, 95 participants (54% male, 55% severe genotype) completed the SMSC assessment. Most participants (87%) transferred to adult care within six months and 87% were retained for at least 12 months. BRIEF and caregiver-reported SMSC assessments were weakly, negatively correlated (ρ=-0.25, p=0.0392) but were not significant in predicting transition outcomes (p>0.05). Conclusions The SMSC and executive function did not predict adult care engagement. Development of readiness assessments that predict care engagement and reflect self-efficacy is important for monitoring transition-aged patients with SCD.

Predictive Analysis of Healthcare Resource Utilization after Elective Spine Surgery.

The management of degenerative spine pathology continues to be a significant source of costs to the US healthcare system. Besides surgery, utilization of healthcare resources after spine surgery drives costs. The responsibility of managing costs is gradually shifting to patients and providers. Patient-centered predictors of healthcare utilization after elective spine surgery may identify targets for cost reduction and value creation. Therefore, our study aims to quantify patterns of healthcare utilization and identify risk factors that predict high healthcare utilization after elective spine surgery. A total of 623 patients who underwent elective spine surgery at a tertiary academic medical center by one of three fellowship-trained orthopedic spine surgeons between 2013 and 2018 were identified in this retrospective cohort study. Healthcare utilization was quantified including advanced spine imaging, emergency and urgent care visits, hospital readmission, reoperation, PT/OT referrals, opioid prescriptions, epidural steroid injections, and pain management referrals. Patient variables, namely, the Charlson comorbidity index (CCI) and the American Society of Anesthesiologists (ASA) classification system, were assessed as potential predictors for healthcare utilization. Among all patients, a wide range of health utilization was identified. Age, body mass index, Charlson Comorbidity Index, and American Society of Anesthesiology class were identified as positive predictors of postoperative healthcare utilization including emergency department visits, spine imaging studies, opioid and nerve blocker prescriptions, inpatient rehabilitation, any referrals, and pain management referrals. Markers of patient health-such as CCI and ASA class-may be used to predict healthcare utilization following elective spine surgery. Identifying at-risk patients and addressing these challenges prior to surgery is an important step to deliver efficient postoperative care. 3.

Personalized Surgical Transfusion Risk Prediction: Reply

We thank Burton et al.1 and Zapf et al.2 for their thoughtful comments on our research on personalized surgical transfusion risk prediction.3 Both letters raise important considerations regarding variable selection for predictive models in health care, which are worth discussing in further detail.As Burton et al. note, race and ethnicity were not included as input variables in our machine learning model for surgical transfusion risk; we would like to clarify that this was intentional for several reasons, which we explain here. First, the inclusion of race in predictive models has been well-described to contribute to inequity.4 One major limitation of machine learning is that a model can only learn from its training examples—in other words, real-world clinician behaviors. If such behaviors or the societal factors contextualizing that behavior are biased, the model will also be biased. The citation provided by Burton et al. is a perfect example of this5: in this study, researchers evaluated a model trained to predict healthcare utilization after hospital discharge, with the intention to allocate additional resources to patients predicted to have high utilization. Unfortunately, black patients had low utilization because they lacked access to care, which the model learned and perpetuated. Inclusion of race as an input variable in model development encourages machine learning models to explicitly encode such latent biases, and consequently the recommendations of such models will propagate systemic inequities in care.Second, although race is a frequently collected variable in many datasets, it serves as a proxy for often unmeasured variables such as socioeconomic status, access to care, illness severity (due to poor access to care and delayed presentation), and other social determinants of health.6 Thus, although the inclusion of race as a variable may improve model discrimination, it potentially does so for the wrong reasons. Given two individuals, identical except for their skin color, it seems unjust for one to have a “better” prediction based on the population averages of their racial group, which may be due to unmeasured variables not applicable to the specific individual.Third, to the best of our knowledge, there is little evidence that race itself contributes to risk for allogeneic blood transfusion after adjustment for disease burden, socioeconomic status, and other clinical variables that are known to contribute (e.g., hematocrit). We thank Burton et al. for bringing attention to the potential pitfalls of racial adjustment and the critical importance of fairness in predictive modeling. As machine learning is increasingly used for clinical decision support, model developers must be vigilant for potential sources of bias, which can be introduced at every step of model development and implementation.7,8 As a research community, we share a responsibility to ensure that the decision support tools we create do not exacerbate, and ideally help to reduce, the health disparities that are currently present in modern medicine.9Zapf et al. raise important points about the benefits and limitations of model development using large registry datasets versus institution-specific datasets. We agree that inclusion of surgeon and anesthesiologist identifiers may further improve predictive performance. Variation in transfusion risk can occur due to differences in surgeon technique or case complexity, and it would be appropriate to adjust for these; however, they can also occur due to differences in preference for discretionary transfusion, which may be less appropriate to adjust for. By training our models on a large national database, we captured the average transfusion behavior of U.S. physicians, which we believe, on average, to be appropriate. Further customizing model predictions based on individual behavior patterns risks encoding undesirable physician practice patterns into the model; nonetheless, we acknowledge that such adjustment might be necessary for widespread adoption. Our transfer learning approach (i.e., hospital-specific procedure-specific transfusion rate) could easily accommodate the addition of a surgeon- or anesthesiologist-specific adjustment, and it would be interesting to investigate such modifications in future work.Dr. Lou was supported by the National Institutes of Health (Bethesda, Maryland; 5T32GM108539-07) and the International Anesthesia Research Society (San Francisco, California).Dr. Kannampallil has consulting relationships with Pfizer Inc. (New York, New York) and Elsevier (Amsterdam, The Netherlands) that are unrelated to this work. Dr. Hall is the Consulting Director of the National Surgical Quality Improvement Program, American College of Surgeons (Chicago, Illinois). The other authors declare no competing interests.

Longitudinal patterns and predictors of healthcare utilization among cancer patients on home-based palliative care in Singapore: a group-based multi-trajectory analysis

BackgroundHome-based palliative care (HPC) is considered to moderate the problem of rising healthcare utilization of cancer patients at end-of-life. Reports however suggest a proportion of HPC patients continue to experience high care intensity. Little is known about differential trajectories of healthcare utilization in patients on HPC. Thus, we aimed to uncover the heterogeneity of healthcare utilization trajectories in HPC patients and identify predictors of each utilization pattern.MethodsThis is a cohort study of adult cancer patients referred by Singapore Health Services to HCA Hospice Service who died between 1st January 2018 and 31st March 2020. We used patient-level data to capture predisposing, enabling, and need factors for healthcare utilization. Group-based multi-trajectory modelling was applied to identify trajectories for healthcare utilization based on the composite outcome of emergency department (ED) visits, hospitalization, and outpatient visits.ResultsA total of 1572 cancer patients received HPC (median age, 71 years; interquartile range, 62–80 years; 51.1% female). We found three distinct trajectory groups: group 1 (31.9% of cohort) with persistently low frequencies of healthcare utilization, group 2 (44.1%) with persistently high frequencies, and group 3 (24.0%) that begin with moderate frequencies, which dropped over the next 9 months before increasing in the last 3 months. Predisposing (age, advance care plan completion, and care preferences), enabling (no medical subsidy, primary decision maker), and need factors (cancer type, comorbidity burden and performance status) were significantly associated with group membership. High symptom needs increased ED visits and hospitalizations in all three groups (ED visits, group 1–3: incidence rate ratio [IRR] 1.74–6.85; hospitalizations, group 1–3: IRR 1.69–6.60). High home visit intensity reduced outpatient visits in all three groups (group 1–3 IRR 0.54–0.84), while it contributed to reduction of ED visits (IRR 0.40; 95% CI 0.25–0.62) and hospitalizations (IRR 0.37; 95% CI 0.24–0.58) in group 2.ConclusionsThis study on HPC patients highlights three healthcare utilization trajectories with implications for targeted interventions. Future efforts could include improving advance care plan completion, supporting care preferences in the community, proactive interventions among symptomatic high-risk patients, and stratification of home visit intensity.

Predictors of Health Care Utilization Following Major Head and Neck Oncologic Surgery in Older Patients.

To identify patient factors in older patients associated with making posttreatment visits in the first year after major head and neck oncologic surgery. Retrospective cohort study. Academic institution. Patients aged ≥60 years who underwent a neck dissection with or without a free flap reconstruction were retrospectively analyzed. Data collected included patient demographics, comorbidities, social variables, perioperative course, and clinical visits. Within a 1-year postoperative period, the 181 patients in our cohort had a mean ± SD 6.37 ± 3.6 postoperative clinic visits; 70% attended at least 4 visits. Multivariable regression analysis showed a significant association with distance closer to the hospital (P = .013): for every 10-mile increase in distance, the number of visits decreased by 0.15 (SE = 0.06). Additionally, receiving adjuvant radiation therapy (P = .0096) demonstrated significant associations: when compared with no adjuvant therapy, radiation therapy had on average 1.5 (SE = 0.56) more visits, and chemoradiation had 0.04 (SE = 0.73) more visits. Older patients who undergo major head and neck oncology surgery are more likely to attend posttreatment visits in the 1 year following surgery if they are discharged home rather than to a skilled nursing facility, live closer to the hospital, and undergo adjuvant radiation therapy.

Engagement in a Social Needs Navigation Program and Health Care Utilization in Pediatric Primary Care.

Evidence for social needs interventions on resource linkage has grown over the past decade. Though social and economic needs predict health care utilization, few studies have assessed social needs interventions on these outcomes. The objective of this study was to examine the association between enrollment in a clinic-based social needs program on subsequent well-child visit (WCV) attendance and emergency department (ED) use in 2 primary care clinics. A retrospective cohort study was conducted of patients in 2 clinics referred to a social risk screening and navigation program between June and August 2018. We compared health care utilization over the subsequent 12 months for those who completed a full intake and received resource navigation (enrolled) to those referred (not enrolled). Logistic regression was used to assess the association between enrollment with WCV attendance and ED visits. During the study period, of the 969 patients referred to the program, 761 were enrolled across the 2 clinics. At both sites, the majority of enrolled patients had high WCV attendance postintervention (Site A: 81.6%, Site B: 71.4%). High WCV attendance for nonenrolled but referred patients was significantly lower (Site A: 52.7%, Site B: 35.0%). Enrolled participants were significantly more likely to have high WCV attendance than nonenrolled patients, adjusting for preintervention utilization (Site A adjusted odds ratio [aOR]: 5.83, Site B aOR : 4.20). There were no significant differences in ED use at either clinic. Addressing families' social needs through resource linkage and navigation can improve WCV attendance.

Health care utilization in young adults with childhood physical disabilities: a nationally representative prospective cohort study

BackgroundYoung people with physical disabilities face barriers to accessing health care; however, few studies have followed adolescents with physical disabilities longitudinally through the transition of care into adulthood. The objective of this study was to investigate differences in health care utilization between adolescents with physical disabilities and those without during the transition period from adolescent to adult care.MethodsWe utilized the National Longitudinal Study of Adolescent to Adult Health, a prospective cohort study following adolescents ages 11–18 at baseline (1994–1995) through adulthood. Baseline physical disability status was defined as difficulty using limbs, using assistive devices or braces, or having an artificial limb; controls met none of these criteria. Health care utilization outcomes were measured seven years after baseline (ages 18–26). These included yearly physical check-ups, unmet health care needs, and utilization of last-resort medical care, such as emergency departments, inpatient hospital wards, and inpatient mental health facilities. Multiple logistic regression models were used to predict health care utilization, controlling for age, sex, race/ethnicity, insurance status, and history of depression.ResultsThirteen thousand four hundred thirty-six participants met inclusion criteria, including 4.2% with a physical disability and 95.8% without. Half (50%) of the sample were women, and the average age at baseline was 15.9 years (SE = 0.12). In logistic regression models, those with a disability had higher odds of unmet health care needs in the past year (Odds Ratio (OR) 1.41 95% CI 1.07–1.87), two or more emergency department visits in the past five years (OR 1.34 95% CI 1.06–1.70), and any hospitalizations in the past five years (OR 1.36 95% CI 1.07–1.72). No statistically significant differences in preventive yearly check-ups or admission to mental health facilities were noted.ConclusionsYoung adults with physical disabilities are at higher risk of having unmet health care needs and using last-resort health care services compared to their non-disabled peers.

Predictors of Healthcare Utilization Among Older Adults with Moderate to Severe Cognitive Limitations in Ghana: A Cross-Sectional Analysis of the WHO Study on Global Ageing and Adult Health (SAGE) Wave 1.

As the number of older adults in Ghana keeps increasing, so does the prevalence of aging-related diseases and conditions that tend to lead to cognitive decline. However, knowledge on the predicting factors of healthcare utilization among older adults with cognitive limitations is still scarce. This study examines the predictive factors of healthcare-seeking behavior among older adults with moderate-to-severe cognitive limitations in Ghana. Based on Andersen's behavioral model of health service utilization, the study analyzed data from 3106 older adults with moderate to severe cognitive limitations in Ghana aged 50+ using Bivariate Probit Regression and the Heckman Selection Model. Results showed that individuals who were aged 75+, living in urban settlements, educated, higher income levels and parents who had some level of education, pension benefits, and both mandatory and voluntary health insurance were more likely to seek inpatient care. Enabling resources such as the national health insurance scheme and other types of medical insurance, household income, pension benefits, and predisposition factors like education influenced outpatient and inpatient healthcare consumption for older adults with moderate to severe cognitive limitations but had little influence on the choice of healthcare facility except for health insurance. Since the healthcare-seeking behavior of older adults with moderate-to-severe cognitive limitations is driven by an amalgamation of several internal and external factors, there is a need for policy change to understand and cautiously incorporate these factors into the decision to improve equitable access to healthcare services for these individuals, as not all sizes fit all in providing quality care for them.