The results of this behavioral intervention study included: the flow of subjects through this study to the analysis of data, demographic characteristics of the sample, the primary outcome results, association of risk perceptions and other variables at pre- and post-intervention with the primary outcome, and the impact of each telephone call on obtaining a DFE.
portrays the flow of participants from the available pool of subjects for screening through study analyses. Of the 10,273 names of adults with diabetes from three healthcare systems in the Bronx, a total of 6045 were contacted and assessed for eligibility by telephone. Of those, 3458 individuals were ineligible, the most frequent reasons were: self-report of having had a DFE during the previous 12 months (60%); self-report of no diabetes (16%); and reported as having died (7%). There were 38 additional subjects (21 in the telephone group and 16 in the print group) who were excluded from the final analyses due to belated discovery of failure to meet an eligibility criterion. That occurred when documentation of a DFE less than 12 months before their randomization was found, which had not been reported during the eligibility assessment. Subjects who withdrew consent or died during the 6-month study timeframe were not excluded from the analysis; rather, their DFE status as of the time of withdrawal or death was used as their outcome result.
Flow of subjects through the “Vision is Precious” study, from eligibility assessment to analysis
Selected Characteristicsacteristics of the Study Sample
These 598 subjects were 60.5% women with a mean age of 56.6 (SD=12.5) years. Forty-five percent reported their race as black and only 17% white; another 27% reported their race as unknown or refused to report it, partly because there was no option to report Hispanic/Latino as a race, as opposed to an ethnic category. Hispanic/Latino ethnicity was reported by 42.5%, with 22.7% requesting Spanish as their preferred language for the study. The annual household income was reported as <$30,000 by 60.4% of the sample, with 38.5% reporting an income <$15,000. There were no significant differences between the two study groups on any characteristics.
In carrying out the telephone intervention, there were 930 completed telephone calls to the intervention group subjects, averaging 8.8 minutes. Subjects received, on average, 3.2 phone calls and spoke with a health educator for 28.1 minutes over the 6-month intervention. In order to accomplish this, it was necessary to attempt an additional 4147 calls that did not connect (e.g., busy signal, wrong number). A full accounting of the costs of the intervention is beyond the scope of this report and is in preparation separately.
The intention-to-treat analysis for the primary outcome of documented receipt of a DFE within 6 months of randomization is shown in with the results of the Solomon 4-group analysis at the top of the table and the combined group results at the bottom. The relative risk (telephone group/print group) of having a DFE within 6 months was 1.73 among those randomized to PRE survey, and 1.74 among those not randomized to PRE survey; these were not significantly different (p=0.97). Therefore, having concluded that completing the PRE survey affected neither the overall probability of having a DFE, nor the effectiveness of the telephone intervention, the main results showed a 74% increase in the probability of screening for diabetic retinopathy in the telephone group compared to the print group (RR=1.74 [95% CI=1.31–2.30], χ2=15.63, p<0.0005} (see ).
Primary outcome results by randomized group in Solomon 4-group design and total combined groups
There were no significant differences in the effect of the intervention by ethnicity (Hispanic, non-Hispanic, and not reported), by language (English or Spanish), or by healthcare site.
Four aspects of the relationships among RPS-DM scale scores, intervention, and study outcome were examined among those 226 subjects who were randomized to the PRE and POST survey group and completed these surveys. Of those randomized to both PRE and POST surveys, 75% completed both. shows that each RPS-DM scale had similar pre-intervention distributions in the telephone and print groups, with no statistically significant differences. The composite risk, comparative disease risk, environmental risk, and worry scales all exhibited statistically significant (p<0.05) changes from pre- to post-intervention in the telephone group. In the print group, worry and risk-knowledge scores changed significantly during the course of the study. The difference between telephone and print groups in the amount of change over time in a scale score was significant at the 0.05 level for the composite risk, personal disease risk, environmental risk, and risk-knowledge scales. The change in worry scores from pre- to post-intervention was essentially the same in both study groups.
Risk perception scale scores at pre- and post-intervention by study group
There were no associations between the change in any of the scale scores and the probability of getting a DFE. The inclusion of these variables in a model did not alter materially the estimate of the treatment effect on DFE outcome. The same conclusions hold if the model is further adjusted for demographic variables.
The only RPS-DM scale whose baseline value mediated the effect of the telephone intervention was the worry scale. The model estimates of the OR for having a DFE within 6 months of randomization (telephone/print) was 3.47 (95% CI=1.78–6.77) for subjects with initial worry score at the group mean of 3.08, whereas it was 1.04 (95% CI=0.31–3.48) for subjects whose pre-intervention worry score was one unit lower (2.08). Thus the telephone intervention was more effective with subjects whose initial level of worry about diabetes was higher. The pre-intervention optimistic bias score, itself, independently predicted DFE outcome, with each unit increase in the score being associated with reducing the odds of obtaining a DFE by a factor of 0.40 (95% CI=0.23–0.73). However, the effectiveness of the telephone intervention was not altered by the optimistic bias score. These results also remain essentially unchanged if adjusted for demographics.
The association of the HbA1c value with the primary study outcome was studied from the perspective of level of metabolic control. Those who had an HbA1c value available during the 12-month timeframe (49.5% [n=145] of the print group and 50.8% [n=155] in the telephone group [p=0.75]) were classified as poorly controlled (HbA1c >9%, n=106), moderately controlled (7%–9%, n=107), or well-controlled (≤7%, n=87). Among those with poor glycemic control, the probability of DFE within 6 months of randomization was only 16.4% in the print group, but 43.1% in the telephone group. For those with moderate glycemic control, the corresponding probabilities were 31.4% and 41.1%, respectively; and for those with good control, 28.2% and 41.7%. The relative risks for having a DFE in the poor, moderate, and good glycemic control groups did not differ.
The effectiveness of each telephone call was estimated by conceptualizing the number of calls as a “time” variable, the occurrence of a DFE as a “death,” and carrying out a “survival” analysis. Subjects who never obtained a DFE were analyzed as requiring an indeterminate, possibly infinite, number of calls exceeding the number actually received. Before the first call, 3.1% (95% CI=1.1–5.1) of subjects obtained a DFE. The initial call propelled 10.4% (95% CI=6.4–14.4) into action. After the fourth, apparently most effective call, 18.1% (95% CI=9.8–26.4) of those who had not yet obtained a DFE did so. Only 9.1% (95% CI=1.1–17.1) did following the fifth. There was then a sharp decline, with the sixth and seventh calls being ineffective.