Table 2:
Prevalence of predictor variables for patients with incident HIV infection versus controls in the development cohort, and LASSO model coefficients for each predictor variable.
| Electronic health record predictor variablesa | Incident HIV (n=150) | Controls (n=7,466) |
Coefficientb |
|---|---|---|---|
| Diagnosis codes, n (%) | |||
| Syphilis of any site or stage except late latent | 6 (4·0) | 5 (0·1) | 1·00 |
| HIV counseling in prior 2 years | 8 (5·3) | 26 (0·3) | 1·10 |
| Contact with or exposure to venereal disease | 15 (10·0) | 139 (1·9) | 0·29 |
| Laboratory tests | |||
| Number of positive gonorrhea tests in prior 2 years, mean (SD) | 0·04 (0·23) | 0·00 (0·02) | 3·07 |
| Number of Chlamydia tests, mean (SD) | 0·00 (0·0) | 0·00 (0·03) | −0·15 |
| Number of HIV tests, mean (SD) | 0·81 (1·71) | 0·18 (0·62) | 0·12 |
| Number of HIV ELISA tests, mean (SD) | 0·61 (1·35) | 0·15 (0·54) | 0·16 |
| Number of HIV tests in prior 2 years, mean (SD) | 0·44 (0·97) | 0·09 (0·34) | 0·23 |
| Number of HIV RNA tests in prior year, mean (SD) | 0·05 (0·40) | 0·00 (0·02) | 0·15 |
| Testing for acute HIVc, n (%) | 7 (4·7) | 7 (0·1) | 1·82 |
| Testing for acute HIVc in prior 2 years, n (%) | 4 (2·7) | 2 (< 0·1) | 0·16 |
| Prescriptions, n (%) | |||
| Intramuscular penicillin G benzathine | 8 (5·3) | 2 (< 0·1) | 1·80 |
| Intramuscular penicillin G benzathine in prior year | 5 (3·3) | 0 (0·0) | 1·36 |
| Intramuscular penicillin G benzathine in prior 2 years | 5 (3·3) | 1 (< 0·1) | 0·21 |
| Buprenorphine and naloxone in prior 2 years | 2 (1·3) | 26 (0·3) | 0·20 |
| Registration data | |||
| Years of prior electronic health records data, mean (SD) | 2·74 (2·72) | 3·92 (2·68) | −0·07 |
| At least 1 year of prior electronic health records data, n (%) | 92 (61·3) | 6153 (82·4) | −0·63 |
| At least 2 years of prior electronic health records data, n (%) | 72 (48·0) | 5230 (70·1) | −0·40 |
| Any data on primary language, n (%) | 129 (86·0) | 7145 (95·7) | −0·08 |
| English as primary language, n (%) | 114 (76·0) | 6778 (90·8) | −0·42 |
| Black race, n (%) | 51 (34·0) | 655 (8·8) | 1·06 |
| White race, n (%) | 55 (36·7) | 5555 (74·4) | −0·66 |
| Male gender, n (%) | 120 (80) | 5967 (79·9) | 1·87 |
ELISA, enzyme-linked immunosorbent assay; RNA, ribonucleic acid.
a
The variables shown are those included in the final LASSO algorithm.
b
To calculate an HIV risk prediction score, the value of each variable is multiplied by its coefficient and the products are then summed to generate the risk score on the logit scale. Binary variables are assigned a value of 1 if affirmative and 0 if non-affirmative.
c
Testing for acute HIV defined as HIV RNA testing among individuals without evidence of HIV infection.