Association of Tolerance of Uncertainty With Outcomes in a Quality Improvement Collaborative Free

The Value in Inpatient Pediatrics (VIP) network is a collaborative network of hospitals organized by the American Academy of Pediatrics (AAP) that supports nationwide implementation of evidence-based guidelines.¹⁰  In 2019, the VIP network chose Standardization of Fluids in Inpatient Settings (SOFI) as its QI project, seeking to increase the adoption of the AAP Clinical Practice Guideline that endorsed the use of isotonic maintenance fluids.¹¹ 

SOFI, as a QI project, had 2 main goals: (1) increasing the time (proportion of hours per hospital day) on exclusive isotonic fluids and (2) decreasing the proportion of hospital days on which laboratory tests were obtained. The project was structured as a stepped-wedge, cluster randomized trial and was conducted across 106 pediatric hospitals in the United States.¹² 

Questions from the Physicians’ Reactions to Uncertainty scale were used to measure tolerance of uncertainty, specifically the subscale examining anxiety due to uncertainty or tolerance of uncertainty as framed by Gerrity et al.¹³  The 5 statements on this scale have a response range from 1 to 6, from strongly disagree to strongly agree. The statements include (1) I usually feel anxious when I am not sure of a diagnosis, (2) I find the uncertainty involved in patient care disconcerting, (3) Uncertainty in patient care makes me uneasy, (4) I am quite comfortable with the uncertainty in patient care, (5) The uncertainty of patient care often troubles me. The items are summed, and an average response taken as the individual’s score. In the original sample of physicians, this scale had a Cronbach’s α of 0.86.

The tolerance of uncertainty subscale was administered to all individual participants at the beginning of the study, after enrollment of their hospital in the project. Participants included attending physicians, nurses, pharmacists, and other providers identified by hospital site leaders as being part of the local implementation team.

SOFI had 2 primary outcomes: (1) mean proportion of hours per hospital day with exclusive isotonic maintenance intravenous fluid (IVF) use and (2) mean proportion of hospital days with laboratories obtained. These were ascertained through chart review at the hospital level and are described in detail separately.¹²  Laboratory testing was measured by the proportion of hospital days on which a white blood cell count was obtained. This laboratory was intentionally chosen for its lack of direct relationship with IVF use to independently measure the impact of our laboratory reduction interventions. Serum sodium laboratory monitoring was monitored separately as a balancing measure, defined as the proportion of hospital days on which a serum sodium was obtained while on maintenance IVF.

We examined the distribution of the tolerance of uncertainty subscale and variation between hospitals and within hospitals (Supplemental Fig 2). We used the mean response of individuals completing the scale at each hospital as a hospital-level estimate given that the outcomes were also at the hospital level. The mean response for each hospital was then scaled to have an overall mean of 0 and standard deviation (SD) of 1 for the analysis. For the effect modification analysis, tolerance of uncertainty was categorized as low, medium, or high based on tertiles of the data.

Exposure to the SOFI intervention was determined by the stepped-wedge, cluster randomized design of the trial.¹²  Each site or hospital was randomized to 1 of 3 implementation start dates. Exposure to SOFI was determined as after these dates at a given site and nonexposure to SOFI before the start date at a given site.

We used linear models to examine the association between exposure to SOFI and the outcomes of time on exclusive isotonic fluids and proportion of days with a laboratory assessment of sodium or white blood count. We chose a priori to examine whether any effect of the SOFI intervention was moderated by the tolerance of uncertainty subscale at the hospital level for each of the 2 aims. The base model included only the SOFI intervention, the second model examined any additional effect of the Reactions to Uncertainty scale, and the interaction model included an interaction term to examine for effect modification. R software (version 4.0.3) was used for all analyses.

For this analysis, we examined responses grouped across 97 of the 106 hospitals that participated, with 423 individual providers contributing responses out of 853 individual e-mail invitations sent, which was based on the people identified by each hospital lead for the project. The number of individuals responding per hospital had a median of 4 and an interquartile ratio of 2 to 5. There was not a significant difference in the mean tolerance of uncertainty by respondent role, although the vast majority of participants were physicians (n = 332, mean 3.4, SD = 0.9), with smaller numbers of nurses (n = 50, mean 3.6, SD = 1.03) and pharmacists (n = 34, mean 3.1, SD = 1.2) participating. The overall mean reported tolerance of uncertainty was 3.39 (95% confidence interval: 3.27–3.50) of a possible range from 1 to 6. The high tolerance group, defined as the top quartile (n = 22 hospitals), had a mean tolerance score of 2.4; the moderate tolerance group, defined as the middle 2 quartiles (n = 50 hospitals), had a mean tolerance score of 3.4, and the low tolerance group (n = 24 hospitals) had a mean tolerance score of 4.2. Supplemental Fig 2 shows the variance within and across all hospitals.

In a multivariable analysis examining the association between exposure to the SOFI intervention and aim 1, time on isotonic fluids, we found that the SOFI intervention was associated with a significant increase in proportion of hours on isotonic fluids, P <.001 (Table 1, model 1), consistent with the previously reported analysis.¹²  When examining for effect modification in model 3 (Table 1), we found that time on isotonic fluids was increased most at lower tolerance of uncertainty (see also Fig 1). Note that the regression analysis shows a significant interaction using a continuous variable and Fig 1 demonstrates the effect modification visually, although it shows overlapping intervals. Hospitals with a lower mean tolerance of uncertainty before the SOFI intervention had markedly lower use of exclusive isotonic fluid before the SOFI intervention (Fig 1), and they experienced the biggest change in practice to increase time on isotonic fluids.

As a sensitivity analysis, we examined only the responses of physicians who completed the survey, and the sample size was reduced by 28%. In this analysis, the same overall pattern was observed in that the group, with the lowest baseline tolerance of uncertainty having the biggest gains in the intervention, although the model did not show a significant interaction effect for tolerance of uncertainty.

In contrast to the findings for aim 1, we did not find evidence of effect modification on the association between exposure to the SOFI intervention and aim 2 (Table 2). Namely, we observed no effect modification of the tolerance of uncertainty scale on the percentage of days with white blood cell counts or serum sodium laboratory tests.

Our findings are the first to examine the association between tolerance of uncertainty and outcomes of a QI initiative. We found that, measured at baseline, less tolerance of uncertainty was associated with greater uptake of isotonic maintenance intravenous fluid use after the SOFI intervention. This group also had lower isotonic fluid use before the SOFI intervention compared with those with moderate and high tolerance of uncertainty, so there was larger ability for improvement in adherence to exclusive isotonic use in those with lower tolerance of uncertainty compared with hospital groups with moderate and high tolerance of uncertainty. This finding contradicted our initial hypothesis in which we expected that a higher tolerance of uncertainty would be associated with increased uptake of the intervention because the routine use of hypotonic maintenance fluids is a common practice dating back to the 1950s and the shift to exclusive isotonic fluids upends that practice. More study is warranted to better characterize how both individual and collective decision-making influence the uptake and effectiveness of QI initiatives.

Contrary to our hypothesis, aim 2, which asked providers to stop ordering laboratory tests, was not associated with the baseline level of tolerance of uncertainty. We had expected that this aim, which was more focused on deimplementation rather than substitution, might be more strongly associated with tolerance of uncertainty. The null findings related to tolerance of uncertainty for aim 2 could be explained by a few possibilities. This aim may have been perceived as a less important piece of the QI project because laboratory testing was not included in the AAP Guideline, does not have a level of evidence associated with it specifically, and was low at baseline, potentially dampening its urgency as an area for improvement. Additional research is necessary to explore providers’ perceived need to check laboratory studies for patients.

No studies were found that analyzed tolerance of uncertainty and its relationship to QI programs. The authors of several studies have explored tolerance of uncertainty and its impact on clinical care, finding increased burnout, reduced career satisfaction, and less engagement at work are associated with lower tolerance of uncertainty.⁴  Other studies have demonstrated that lower tolerance of uncertainty can lead to unnecessary testing and iatrogenic injury; however, we found no correlation between varying levels of tolerance of uncertainty and increased testing in the SOFI initiative.¹⁴  Another study explored tolerance of uncertainty in emergency medicine providers, with findings suggesting that physicians have bias in how much uncertainty patients are willing to tolerate, which increases the number of unnecessary tests and/or treatments.¹⁵  Greater clinical experience has been found to be associated with lower risk aversion strategies (eg, more testing, inpatient admission).¹⁶  Such strategies flow from lower tolerance of uncertainty.

Our findings imply that tolerance of uncertainty warrants further study as a contextual factor in QI. The SOFI interventions did not explicitly target tolerance of uncertainty as an area for change, and further research in this area could inform how QI interventions are designed and adapted. These results highlighted a larger gap between the baseline and target outcome for the groups with lowest tolerance of uncertainty, showing the importance of understanding context in large QI initiatives.^17,18,19  Previous studies have found an increased failure to follow evidenced-based guidelines secondary to one’s tolerance of uncertainty,²⁰  which may help explain why some QI initiatives are more successful than others. Interestingly, our study found lower tolerance of uncertainty was associated with higher levels of uptake in isotonic fluids after the SOFI QI initiative; however, laboratory draws still occurred independent of tolerance of uncertainty. Providers who are less tolerant of uncertainty can increase medical costs because of excessive care, which the SOFI study illustrates with excessive blood draws. The results from this do not allow us to differentiate between the possibilities that tolerance of uncertainty is a surrogate for inertia or that it influences uptake of the QI initiative. Further research that examines tolerance of uncertainty in the context of QI initiatives and follows it over time could help differentiate these.

The study has several limitations, including using individual providers’ tolerance of uncertainty to create group means and the association with patients’ outcomes in which we do not know who provided care for that patient. The providers who reported their tolerance of uncertainty in the QI project are a select sample of each hospital’s providers and may not represent the hospital-wide tolerance of uncertainty. If anything, this selection bias probably skews the results toward greater tolerance of uncertainty, in comparison with all of the hospital staff, because those involved in QI projects may be more engaged in change. We assume that providers’ tolerance of uncertainty was stable over time during the intervention because we only measured it at baseline, which may not be true.

This novel study analyzed the results of a multicenter quality improvement initiative aimed at increasing exclusive isotonic fluid use in otherwise healthy pediatric patients admitted to the hospital and rates of laboratory draws while on isotonic fluids in relation to levels of tolerance of uncertainty, measured at baseline. We found that initial lower tolerance of uncertainty was associated with increase in exclusive isotonic fluid use after the SOFI QI initiative but was not associated with a reduction in laboratory draws. Our results suggest that tolerance of uncertainty as a contextual factor is worth further investigation to explore how it may interact with other individual and systemic factors to affect QI performance.