Setting / Database
This study used the 2017 NIS database, available through the Agency for Healthcare Research and Quality provided by the Healthcare Cost and Utilization Project (7). NIS is the largest United States based publicly available all-payer inpatient health care database (7). It approximates a 20% stratified sample of US hospital discharges from 46 participating states. The NIS has data for more than 7 million unweighted hospital stays per year. When weighted to represent all admissions, it estimates more than 35 million hospitalizations annually, and represents about 95% of US hospitalizations. Strata include hospital size/volume, teaching status, geographic region, and hospital ownership. Data from 2017 NIS uses the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) coding system for all discharge diagnoses.
Study Population, Patient and Hospital Characteristics, and Outcomes
All patients ≥18 years of age were included in the sample. We then identified selected patients with bilateral severe visual impairment or bilateral blindness as described by their ICD-10 CM codes (eye category 2 through 5 for either eye): https://www.icd10monitor.com/looking-at-new-icd-10-cm-codes-for-blindness (updated September 27th 2017). In addition, ICD-10 code H54.0 was also used for bilateral blindness. Supplementary Table A with ICD-10 codes provides specific descriptions of categories for each level of SVI/B. Severe blindness has been defined as individuals with visual acuity worse than 6/60, and blindness as those with visual acuity worse than 3/60 (8).
ICD-10 codes used for our subgroup analysis to study the impact of obesity as a secondary diagnosis on patients with SVL/B was also retrieved (See Supplementary A table for ICD-10 codes and corresponding diagnoses). For adults, obesity is defined as having a Body Mass Index (BMI) of 30 or greater (9). BMI is calculated by taking the individual's weight in kilograms and dividing it by their height in meters squared.
Data was collected and adjusted for select patient and hospital characteristics including age, gender (male and female only), race (White, Black, Hispanic, Asian or Pacific Islander), insurance (Medicare, Medicaid, Private Insurance, Uninsured), median household income (1. $1-$38,999 2. $39,000-$47,999 3. $48,000-$62,999 4. $63,000 or more), based on home zip code, and the Charlson comorbidity index (CCI: score 0 = no comorbidities score 1= low comorbidity burden, score 2 = moderate comorbidity burden, and score 3 or greater = high comorbidity burden). The CCI has been used extensively in clinical research; it is commonly used to assess mortality risk and it is supported by extensive validity evidence (10). Higher scores have been associated with mortality or greater healthcare resource use (11).
The primary outcome was mortality during hospitalization; secondary outcomes were total hospital charges, length of stay (LOS), and disposition after hospitalization. Disposition indicates the discharge location or where patients go after hospitalization. This is most often home, but not infrequently can be elsewhere including venues such as other hospitals, inpatient hospice, inpatient rehabilitation facilities, and nursing homes (https://www.hcup-us.ahrq.gov/db/vars/sedddistnote.jsp?var=dispuniform).
Our Institutional Review Board designated this work as being exempt from detailed review (IRB review number: 00257552).
Statistical analyses
Comparisons were examined between patients with SVI/B and the general population without visual impairments using Pearson’s χ2 tests and one-way analysis of variance to test categorical and continuous variables. Analyses were also carried out within the SVI/B patient cohort assessing those with and without obesity. The primary and secondary outcomes were adjusted for all of the patient demographics and hospital characteristics shown in Table 1, as well as the CCI and select specific comorbidities described in Table 2.
Adjusted odds ratios [aOR] and adjusted mean differences [aMD] from multivariate logistic and linear regression analyses were obtained. Binary outcomes under logistic regression analyses (in-hospital mortality and discharge disposition) were studied. Linear regression was used to study continuous outcome variables (including total hospital charges and LOS). Stata 15.0 statistical software (Stata Corp, College Station, TX) was used and permitted us to account for design complexity (stratification, weighting, and clustering) (7). The p-values for this study were 2 sided and type I error significance level was set at 0.05.