Use of the ICD-10 vision codes to study ocular conditions in Medicare beneficiaries with stroke

Machine Learning Methods Using Artificial Intelligence Deployed on Electronic Health Record Data for Identification and Referral of At-Risk Patients From Primary Care Physicians to Eye Care Specialists: Retrospective, Case-Controlled Study

BACKGROUND

Identification and referral of at-risk patients from primary care practitioners (PCPs) to eye care professionals remain a challenge. Approximately 1.9 million Americans suffer from vision loss as a result of undiagnosed or untreated ophthalmic conditions. In ophthalmology, artificial intelligence (AI) is used to predict glaucoma progression, recognize diabetic retinopathy (DR), and classify ocular tumors; however, AI has not yet been used to triage primary care patients for ophthalmology referral.

OBJECTIVE

This study aimed to build and compare machine learning (ML) methods, applicable to electronic health records (EHRs) of PCPs, capable of triaging patients for referral to eye care specialists.

METHODS

Accessing the Optum deidentified EHR data set, 743,039 patients with 5 leading vision conditions (age-related macular degeneration [AMD], visually significant cataract, DR, glaucoma, or ocular surface disease [OSD]) were exact-matched on age and gender to 743,039 controls without eye conditions. Between 142 and 182 non-ophthalmic parameters per patient were input into 5 ML methods: generalized linear model, L1-regularized logistic regression, random forest, Extreme Gradient Boosting (XGBoost), and J48 decision tree. Model performance was compared for each pathology to select the most predictive algorithm. The area under the curve (AUC) was assessed for all algorithms for each outcome.

RESULTS

XGBoost demonstrated the best performance, showing, respectively, a prediction accuracy and an AUC of 78.6% (95% CI 78.3%-78.9%) and 0.878 for visually significant cataract, 77.4% (95% CI 76.7%-78.1%) and 0.858 for exudative AMD, 79.2% (95% CI 78.8%-79.6%) and 0.879 for nonexudative AMD, 72.2% (95% CI 69.9%-74.5%) and 0.803 for OSD requiring medication, 70.8% (95% CI 70.5%-71.1%) and 0.785 for glaucoma, 85.0% (95% CI 84.2%-85.8%) and 0.924 for type 1 nonproliferative diabetic retinopathy (NPDR), 82.2% (95% CI 80.4%-84.0%) and 0.911 for type 1 proliferative diabetic retinopathy (PDR), 81.3% (95% CI 81.0%-81.6%) and 0.891 for type 2 NPDR, and 82.1% (95% CI 81.3%-82.9%) and 0.900 for type 2 PDR.

CONCLUSIONS

The 5 ML methods deployed were able to successfully identify patients with elevated odds ratios (ORs), thus capable of patient triage, for ocular pathology ranging from 2.4 (95% CI 2.4-2.5) for glaucoma to 5.7 (95% CI 5.0-6.4) for type 1 NPDR, with an average OR of 3.9. The application of these models could enable PCPs to better identify and triage patients at risk for treatable ophthalmic pathology. Early identification of patients with unrecognized sight-threatening conditions may lead to earlier treatment and a reduced economic burden. More importantly, such triage may improve patients' lives.

A Spatial Variation Analysis of In-Hospital Stroke Mortality Based on Integrated Pre-Hospital and Hospital Data in Mashhad, Iran

BACKGROUND

Despite significant advances in the quality and delivery of specialized stroke care, there still persist remarkable spatial variations in emergency medical services (EMS) transport delays, stroke incidence, and its outcomes. Therefore, it is very important to investigate the possible geographical variations of in-hospital stroke mortality and to identify its associated factors.

METHODS

This historical cohort study included suspected stroke cases transferred to Ghaem Hospital of Mashhad by the EMS from March 2018 to March 2019. Using emergency mission IDs, the pre-hospital emergency data were integrated with the patient medical records in the hospital. We used the Bayesian approach for estimating the model parameters.

RESULTS

Out of 301 patients (142 (47.2%) females vs. 159 (52.8%) males) with a final diagnosis of stroke, 61 (20.3%) cases had in-hospital mortality. Results from Bayesian spatial log-logistic proportional odds (PO) model showed that age (PO=1.07), access rate to EMS (PO=0.78), arrival time (evening shift vs. day shift, PO=0.09), and sequelae variables (PO=9.20) had a significant association with the odds of in-hospital stroke mortality (P<0.05). Furthermore, the odds of in-hospital stroke mortality were higher in central urban areas compared to suburban areas.

CONCLUSION

Marked regional variations were found in the odds of in-hospital stroke mortality in Mashhad. There was a direct association between age and odds of in-hospital stroke mortality. Hence, the prognosis of in-hospital stroke mortality could be improved by better control of hypertension, prevention of the occurrence of sequelae, increasing the access rate to EMS, and optimizing shift work schedule.

Prevalence and Risk Factors of Glaucoma Among Chinese People.From the China Health and Retirement Longitudinal Study

PRCIS

This study demonstrated the prevalence of self-reported glaucoma and its strong association with pre-existing systemic chronic diseases in China using the baseline data from CHARLS, a nationwide population-based cohort.

PURPOSE

To estimate the prevalence of self-reported glaucoma and its risk factors using data from the China Health and Retirement Longitudinal Study (CHARLS).

MATERIALS AND METHODS

Data on age, sex, area of residence, education, marital status, health-related behaviors, and pre-existing comorbidities for this cross-sectional study were retrieved from the CHARLS for 17,713 subjects who completed a questionnaire between June 2011 and March 2012. The prevalence of glaucoma was estimated, and a multivariate weighted analysis was performed to estimate the odds ratios (ORs) of its risk factors.

RESULTS

Of 16,599 respondents (93.7%) who answered questions regarding glaucoma and their history of systemic chronic diseases, 314 (1.89%) reported having glaucoma before the index date. Qinghai and Beijing had the highest prevalence of glaucoma in China. Glaucoma was significantly associated with hypertension (OR 1.362 [95% confidence interval (CI) 1.801-2.470]), diabetes (OR 2.597 [95% CI 1.661-10.207]), dyslipidemia (OR 1.757 [95% CI 1.157-3.650]), lung disease (OR 2.098 [95% CI 1.674-6.527]), stroke (OR 5.278 [95% CI 1.094-25.462]), heart disease (OR 1.893 [95% CI 1.237-3.363]), and health-related behaviors such as smoking and alcohol consumption after adjusting for age, sex, area, education, marital status, and medical insurance.

CONCLUSIONS

Geographic variation in the prevalence of self-reported glaucoma and its strong association with pre-existing systemic chronic diseases were observed, suggesting that in addition to ophthalmological examinations, regular physical examinations are necessary for glaucoma patients, especially in areas of high incidence. Appropriate strategies to improve preventive measures for glaucoma are recommended for the Chinese population.

Association between vision impairment and cognitive decline in older adults with stroke: Health and Retirement Study

Stroke survivors may experience multiple residual symptoms post-stroke, including vision impairment (VI) and cognitive decline. Prior studies have shown that VI is associated with cognitive decline, but have not evaluated the contribution of VI to post-stroke cognitive changes. We used data from four waves (2010-2016) of the Health and Retirement Study to investigate the cognitive trajectories of stroke survivors with and without VI. Vision (excellent-very good[ref], good, fair-poor) and stroke diagnosis were self-reported. Cognition was defined using the Telephone Interview for Cognitive Status. Regression was used to model the association between vision and change in cognitive function, adjusting for confounders. The final sample included 1,439 stroke survivors and the average follow-up time was 4.1 years. Fair-poor overall (B = -1.30, p < 0.01), near (B = -1.53, p < 0.001), and distance (B = -1.27, p < 0.001) vision were associated with significantly lower baseline cognitive function. VI was not associated with the rate of cognitive decline. Future research should determine whether specific types of VI potentiate the risk of cognitive impairment and dementia in stroke survivors.

Barriers and facilitators to the implementation of a structured visual assessment after stroke in municipal health care services

Background

Stroke is a leading cause of disability worldwide. Visual impairments (VIs) affect 60% of stroke survivors, and have negative consequences for rehabilitation and post-stroke life. VIs after stroke are often overlooked and undertreated due to lack of structured routines for visual care after stroke. This study aims to identify and assess barriers and facilitators to the implementation of structured visual assessment after stroke in municipal health care services. The study is part of a larger knowledge translation project.

Methods

Eleven leaders and municipal interdisciplinary health care professionals participated in qualitative interviews. During two workshops, results from the interviews were discussed with 26 participants from municipal health care services and user representatives. Data from interviews and workshops were collected before the intervention was implemented and analyzed using content analysis.

Results

The analysis identified individual and contextual barriers and facilitators. The individual barriers were related to the participants' experiences of having low competence of visual functions and vision assessment skills. They considered themselves as generalists, not stroke experts, and some were reluctant of change because of previous experiences of unsuccessful implementation projects. Individual facilitators were strong beliefs that including vision in stroke care would improve health care services. If experienced as useful and evidence based, the new vision routine would implement easier. Contextual barriers were experiences of unclear responsibility for vision care, lack of structured interdisciplinary collaboration and lack of formal stroke routines. Time constraints and practical difficulties with including the vision tool in current medical records were also expressed barriers. Contextual facilitators were leader support and acknowledgement, in addition to having a flexible work schedule.

Conclusions

This study shows that improving competence about VIs after stroke and skills in assessing visual functions are particularly important to consider when planning implementation of new vision routines in municipal health care services. Increased knowledge about the consequences of living with VIs after stroke, and the motivation to provide best possible care, were individual facilitators for changing clinical practice. Involving knowledge users, solutions for integrating new knowledge in existing routines, along with easily accessible supervision in own practise, are essential facilitators for promoting a successful implementation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-021-06467-4.

Post-acute care use patterns among Hospital Service Areas by older adults in the United States: a cross-sectional study

BACKGROUND

Despite the success of stroke rehabilitation services, differences in service utilization exist. Some patients with stroke may travel across regions to receive necessary care prescribed by their physician. It is unknown how availability and combinations of post-acute care facilities in local healthcare markets influence use patterns. We present the distribution of skilled nursing, inpatient rehabilitation, and long-term care hospital services across Hospital Service Areas among a national stroke cohort, and we describe drivers of post-acute care service use.

METHODS

We extracted data from 2013 to 2014 of a national stroke cohort using Medicare beneficiaries (174,498 total records across 3232 Hospital Service Areas). Patients' ZIP code of residence was linked to the facility ZIP code where care was received. If the patient did not live in the Hospital Service Area where they received care, they were considered a "traveler". We performed multivariable logistic regression to regress traveling status on the care combinations available where the patient lived.

RESULTS

Although 73.4% of all Hospital Service Areas were skilled nursing-only, only 23.5% of all patients received care in skilled nursing-only Hospital Service Areas; 40.8% of all patients received care in Hospital Service Areas with only inpatient rehabilitation and skilled nursing, which represented only 18.2% of all Hospital Service Areas. Thirty-five percent of patients traveled to a different Hospital Service Area from where they lived. Regarding "travelers," for those living in a skilled nursing-only Hospital Service Area, 49.9% traveled for care to Hospital Service Areas with only inpatient rehabilitation and skilled nursing. Patients living in skilled nursing-only Hospital Service Areas had more than five times higher odds of traveling compared to those living in Hospital Service Areas with all three facilities.

CONCLUSIONS

Geographically, the vast majority of Hospital Service Areas in the United States that provided rehabilitation services for stroke survivors were skilled nursing-only. However, only about one-third lived in skilled nursing-only Hospital Service Areas; over 35% traveled to receive care. Geographic variation exists in post-acute care; this study provides a foundation to better quantify its drivers. This study presents previously undescribed drivers of variation in post-acute care service utilization among Medicare beneficiaries-the "traveler effect".