PREDICTING CONTEXT-SENSITIVITY OF BEHAVIOR IN FIELD DATA
3.3 Study 2: Hand Washing among Hospital Workers Human Subjects ProtectionsHuman Subjects Protections
3.3 Study 2: Hand Washing among Hospital Workers
Table 3.4 provides summary statistics about the workers in our analytic sample. The mean compliance with handwashing is 0.45 per opportunity. An average of 116 shifts are recorded per healthcare worker, and there are an average of 26 episodes (or visits to patient rooms, each with two opportunities to wash - one upon entry and one upon exit) per shift. We observe an average of 3,016 episodes per worker.
Table 3.4: Summary statistics onn= 3,124 hospital caregivers’ hand washing
Mean SD Q1 Median Q3
Hand sanitizing compliance 0.45 0.23 0.26 0.43 0.63
Total number of shifts 116 77 56 98 153
Number of rooms visited 37 33 20 29 41
Avg. episode length (mins) 5.66 2.61 3.94 5.13 6.78
Avg. number of episodes per shift 25.72 16.49 13.95 24.2 34.54 Avg. shift length (mins) 511.91 213.75 408.38 581.2 645.74 Avg time between episodes (mins) 22.42 11.5 13.95 20.12 29.00 Avg time off between shifts (hours) 91.95 57.91 60.06 72.61 102.91
Analytic Approach
We use the same machine learning approach as described in Study 1, training a LASSO model to obtain person-specific sets of coefficients and predictability measurements (AUCs). We obtained an 𝑅2 > 0.5 for 33% of individuals. Model fit was not related to the rate of hand washing compliance. However the well-fit group tended to have a slightly higher number of shifts (116 vs 115, 𝑝 <0.001, Cohen’s d=0.010) and less time off between shifts (89.6 vs 93.1 hours, 𝑝 <0.001, Cohen’s d=0.062) (see Appendix C for further discussion). These differences are highly significant given our statistical power but small in magnitude.
As in Study 1, we inferred habit formation time only for those individuals we could fit in the data, assuming others never became habituated to hand sanitizing. We adopted the same approach of fitting what we call pre- and post-habit data using lengthening pre-habit time intervals. Rather than adding two weeks of attendance data at a time to the pre-habit interval, we added two additional shifts at a time to the growing window blending pre- and post-habit data (up to, on average, 51 episodes).
Results
The LASSO model does a satisfactory job fitting hospital caregivers’ hand washing behavior. In the training dataset, the mean (median) individual-level AUC is 0.788 (0.783), and the interquartile range is 0.742-0.828. In the test dataset these measures
are again lower as would be expected, with a mean (median) individual-level AUC of 0.781 (0.776), and interquartile range of 0.732-0.825. While our LASSO models have slightly less predictive power in this domain (compared to gym attendance), they still outperform random chance at predicting hospital caregivers’ hand washing behavior.
As in Study 1, the AUC measure - which can be used in any behavioral domain - is produced for each individual, and it once again serves as an objective measure of context-sensitive predictability. Furthermore, PCS is again able to narrow down the set of context variables that are the most important predictors of hand washing at the aggregate level (see Table 3.5).
The most important context variable is handwashing compliance during their last shift. Surprisingly, a room entry indicator is negative for 77%.
Time of day intervals were not selected by the LASSO model as predictive of most people’s hand washing behavior. However, consistent with previous research (Dai et al, 2015), the amount of time since the start of a caregiver’s shift was a negative predictor of hand washing for 42% of the caregivers. The most important and homogenous predictors were a hospital worker’s handwashing compliance during their last shift (a positive predictor for 100% of the hospital workers), room entry (which is a negative predictor for 77% of hospital workers, indicating most are more likely to wash their hands upon exiting, rather than entering, a room), and the room compliance of others (a positive predictor for 66% of hospital workers). The most heterogeneous predictors are room frequency (the rate at which a specific room is visited by the hospital worker, compared to other rooms) and time off work (that is, time off between shifts), both of which were equally likely to be positive or negative predictors of hand washing when predictive at all.
Table 3.5: Context predictors of hospital hand washing
Homogeneity
Variable % % % index
importance Q1 Median Q3 zero positive negative (|% pos. - % neg.|)
Compliance last shift 0.77 0.66 0.70 0.92 0 100 0 100
Entry indicator 0.35 -0.33 -0.28 -0.04 18 5 77 72
Compliance last opp.×Entry indicator
0.13 0.00 0.00 0.21 49 47 4 43
Compliance last opp.×Time since last opp.
0.12 0.00 0.00 0.00 54 1 45 44
Compliance within a room 0.12 0.00 0.01 0.14 33 51 16 35
Time since last opp. 0.09 0.00 0.00 0.00 61 24 15 9
(Time since last opp.)2 0.08 0.00 0.00 0.00 74 7 18 11
Room compliance of others 0.08 0.04 0.05 0.12 32 66 2 64
Time at work 0.08 0.00 0.00 0.00 54 4 42 38
Compliance last opp.×(Time since last opp.)2
0.07 0.00 0.00 0.00 74 20 5 15
Prev. room compliance 0.07 0.03 0.04 0.11 32 65 2 63
Compliance last opp. 0.05 0.00 0.00 0.07 47 45 7 38
Time at work×6am-12pm 0.05 0.00 0.00 0.00 78 10 12 2
Time since last compliance 0.05 0.00 0.00 0.00 64 9 27 18
Time at work×12pm-6pm 0.04 0.00 0.00 0.00 73 10 17 7
(Time since last compliance)2 0.03 0.00 0.00 0.00 75 17 8 9
12am-6am 0.03 0.00 0.00 0.00 68 22 10 12
Frequency of patient encounter
0.03 0.00 0.00 0.01 58 31 12 19
Time at work×Patient encounter
0.03 0.00 0.00 0.00 64 8 28 20
Days since start 0.02 0.00 0.00 0.00 83 9 8 1
6am-12pm 0.02 0.00 0.00 0.00 80 7 13 6
12pm-6pm 0.02 0.00 0.00 0.00 77 12 11 1
Room frequency 0.02 0.00 0.00 0.00 63 19 19 0
Time at work×6pm-12am 0.02 0.00 0.00 0.00 82 10 7 3
(Time off)2 0.01 0.00 0.00 0.00 84 8 8 0
October 0.01 0.00 0.00 0.00 81 10 9 1
November 0.01 0.00 0.00 0.00 82 10 8 2
December 0.01 0.00 0.00 0.00 81 10 9 1
March 0.01 0.00 0.00 0.00 82 9 10 1
April 0.01 0.00 0.00 0.00 80 10 11 1
May 0.01 0.00 0.00 0.00 80 9 10 1
June 0.01 0.00 0.00 0.00 80 10 10 0
July 0.01 0.00 0.00 0.00 79 11 10 1
August 0.01 0.00 0.00 0.00 78 11 11 0
September 0.01 0.00 0.00 0.00 82 9 9 0
Day-of-week frequency 0.01 0.00 0.00 0.00 77 13 10 3
Rooms visited in shift 0.01 0.00 0.00 0.00 83 8 9 1
6pm-12am 0.01 0.00 0.00 0.00 84 7 8 1
Prev. day-of-week compliance
0.01 0.00 0.00 0.00 79 8 13 5
Prev. unit compliance 0.01 0.00 0.00 0.00 78 10 13 3
Streak 0.01 0.00 0.00 0.00 78 8 15 7
Time off 0.01 0.00 0.00 0.00 85 8 7 1
Unit frequency 0.01 0.00 0.00 0.00 72 20 8 12
February 0.00 0.00 0.00 0.00 82 8 9 1
As in Study 1, we fit an exponential model to individuals’ AUC sequences and plot these increasingly wider ranges of AUC values for individuals who were well fit by the model. This allows us to analyze the development of predictability over time (Figure 3.3), which in turn serves as a proxy for the speed of habit formation, as defined by reaching 95% of their asymptote of predictability. We see that the median habit is formed on the order of weeks, unlike gym attendance where the median time to habit formation was on the order of months.
Figure 3.3: Development of habit formation of hospital hand washing.
(a) An example of one individual’s hand washing behavior, where habit formation is again modelled as an asymptotically increasing sequence of AUCs over time (orange curve A(t)) and the instanta- neous strength of habit formation is the derivative of𝑡 𝐴(𝑡)with respect to time (green curve D(t)).
The time to habit formation is determined by when the individual reaches 95% of the asymptote and is marked by the blue line. (b) A summary of results from fitting the exponential curves to hospital workers’ AUC sequences. The median estimated time to reach the 95% asymptote across all hospital workers well fit by the exponential model is 14 shifts (interquartile range is 5-37 shifts).
Time to habit formation
0.6 0.7 0.8
0 50 100 150
Shifts since start of observed period
AUC
Fitted A Fitted D
0 50 100 150
0 50 100 150
Time to 95% asymptote Median
Additional Analysis of Reward Sensitivity
As in Study 1, we use individual-level habit formation time estimates to test whether post-habit behavior is less sensitive to a reward change. The hypothesized reward change is the last opportunity a caregiver has to wash their hands in the final room visit for their shift. The hypothesis is that they are less likely to wash their hands because it is less important to do so, for hygiene, when they are leaving. However, there is again no statistically significant effect (see Appendix D).