Effect of Electronic Health Record Clinical Decision Support on Contextualization of Care

Factors identified through probing were much more likely to be incorporated into care plans than those that were not (aOR, 61.65 [95% CI, 21.92-173.42]; P < .001).

“As in past studies, factors identified through probing were additionally much more likely to be incorporated into care plans than those that were not (aOR, 61.65 [95% CI, 21.92-173.42]; P < .001) (Table 1 in eAppendix 6 in Supplement 2).17”

• cites [17] Schwartz A, Weiner SJ, Binns-Calvey A, Weaver FM. Providers contextualise care more often when they discover patient context by asking: meta-analysis of three primary data sets. BMJ Qual Saf. 2016; 25(3): 159- 163. doi: 10.1136/bmjqs-2015-004283 26202131 sourceSupported

The capacity of contextualized CDS to change outcomes is limited, to a varying degree, by the resources available to help patients and the contextual factor.

“The capacity of contextualized CDS to change outcomes is limited, to a varying degree, by the resources available to help patients and the contextual factor.”

• cites [20] Binns-Calvey AE, Malhiot A, Kostovich CT,. Validating domains of patient contextual factors essential to preventing contextual errors: a qualitative study conducted at Chicago Area Veterans Health Administration sites. Acad Med. 2017; 92(9): 1287- 1293. doi: 10.1097/ACM.0000000000001659 28353498 sourceCould not verify

  The Interpersonal/social support section (p. 2) asks how relationships and transportation affect patients’ health management, and Technology/Resources (p. 3) asks about technology use. Searches for contextualized clinical decision support, outcome changes, and limits imposed by available resources found no passage addressing the claimed relationship. (LLM, conf=0.70)

  Experimental — accuracy not yet measured

Contextualizing care is associated with partial or complete resolution of the presenting contextual red flag, a meaningful health care outcome.

“Contextualizing care is clinically important because it is associated with partial or complete resolution of the presenting contextual red flag, a meaningful health care outcome.5”

• cites [5] Weiner S, Schwartz A, Altman L,. Evaluation of a patient-collected audio audit and feedback quality improvement program on clinician attention to patient life context and health care costs in the Veterans Affairs health care system. JAMA Netw Open. 2020; 3(7): e209644. doi: 10.1001/jamanetworkopen.2020.9644 32735338 PMC7395234 sourceSupported

Efforts to improve contextualization of care through audit and feedback have achieved measurable but limited success.

“Efforts to improve contextualization of care through audit and feedback have achieved measurable but limited success.”

• cites [5] Weiner S, Schwartz A, Altman L,. Evaluation of a patient-collected audio audit and feedback quality improvement program on clinician attention to patient life context and health care costs in the Veterans Affairs health care system. JAMA Netw Open. 2020; 3(7): e209644. doi: 10.1001/jamanetworkopen.2020.9644 32735338 PMC7395234 sourceSupported

Contextualized CDS can guide contextually informed decision making by drawing on data specific to the life circumstances and behaviors of individual patients.

“Just as CDS can guide biomedically focused decision-making by drawing on data from evidence-based guidelines and other forms of research evidence, it can also guide contextually informed decision making by drawing on data specific to the life circumstances and behaviors of individual patients.18”

• cites [18] Sutton RT, Pincock D, Baumgart DC, Sadowski DC, Fedorak RN, Kroeker KI. An overview of clinical decision support systems: benefits, risks, and strategies for success. NPJ Digit Med. 2020; 3: 17. doi: 10.1038/s41746-020-0221-y 32047862 PMC7005290 sourcePartly supported

  The Introduction says CDSS matches individual patient characteristics to a knowledge base to produce patient-specific recommendations. The Patient-facing decision support section describes records that can collect insurance coverage and patient-entered information, while the Operational impact section mentions collecting medication-adherence data. These passages support the use of individualized data, but do not establish that CDS uses patients’ life circumstances and behaviors to guide contextually informed decisions. (LLM, conf=0.82)

  Experimental — accuracy not yet measured

When the EHR populated the CCB or activated alerts for red flags before the visit, the likelihood of the clinician probing the red flag increased significantly (aOR, 3.64 [95% CI, 1.18-11.20]; P = .02).

“When the EHR populated the CCB or activated alerts for red flags before the start of the visit (ie, from the patient questionnaire or EHR triggers), the likelihood of the clinician probing the red flag increased significantly (aOR, 3.64 [95% CI, 1.18-11.20]; P = .02)”

Customized clinical decision support (CDS) tools did not improve patient outcomes in this randomized clinical trial.

“In this randomized clinical trial that included 452 audio-recorded patient encounters, customized CDS tools did not improve patients’ outcomes but significantly improved contextualization of their care.”

• Also stated in: § Outcomes · plain

  “For the primary end point, the study group did not have additional significant impact on outcomes beyond its impact on plans (adjusted odds ratio [aOR] for improvement, 0.96 [95% CI, 0.57-1.63; P = .88]; aOR for worsening, 1.31 [95% CI, 0.84-2.04; P = .86) (Table 4).”

• Also stated in: § Discussion · plain

  “In this randomized clinical trial, CDS tools did not improve patient outcomes; however, these tools increased the likelihood that a physician would address relevant patient life context in their care plan.”

Factors selected by patients on the previsit questionnaire were not more likely to be incorporated into care plans (aOR, 1.42 [95% CI, 0.57-3.54]; P = .45).

“but factors selected by patients on the previsit questionnaire were not (aOR, 1.42 [95% CI, 0.57-3.54]; P = .45).”

Red flags selected on the previsit questionnaire were less likely to be probed than those expressed orally during the visit (aOR, 0.15 [95% CI, 0.07-0.30]; P < .001).

“In addition, red flags selected on the previsit questionnaire were less likely to be probed than those expressed orally during the visit (aOR, 0.15 [95% CI, 0.07-0.30]; P < .001) (Table 1 in eAppendix 6 in Supplement 2).”

Factors spontaneously revealed by patients were more likely to be incorporated into care plans (aOR, 8.55 [95% CI, 3.98-18.36]; P < .001).

“Factors spontaneously revealed by patients were also more likely to be incorporated (aOR, 8.55 [95% CI, 3.98-18.36]; P < .001)”

Physicians in the intervention group were more likely to probe red flags than physicians in the control group (aOR, 2.12 [95% CI, 1.14-3.93]; P = .02).

“Physicians in the intervention group were more likely to probe red flags than physicians in the control group (aOR, 2.12 [95% CI, 1.14-3.93]; P = .02) (Table 4).”

Customized CDS tools significantly improved contextualization of care, measured as clinician incorporation of contextual factors into the care plan.

“In this randomized clinical trial that included 452 audio-recorded patient encounters, customized CDS tools did not improve patients’ outcomes but significantly improved contextualization of their care.”

• Also stated in: § Outcomes · plain

  “Factors in the intervention group had a higher likelihood of incorporation into the care plan (aOR, 2.67 [95% CI, 1.32-5.41]; P = .006) (Table 4).”

• Also stated in: § Discussion · plain

  “In this randomized clinical trial, CDS tools did not improve patient outcomes; however, these tools increased the likelihood that a physician would address relevant patient life context in their care plan.”

There was no effect of whether the recorder was visible or concealed on probing (aOR, 1.57 [95% CI, 0.63-3.96]; P = .34).

“There was no effect of whether the recorder was visible or concealed (aOR, 1.57 [95% CI, 0.63-3.96]; P = .34).”

When a factor was present, the likelihood of incorporating it into the care plan increased when the EHR populated the CCB or activated alerts (aOR, 11.27 [95% CI, 2.27-55.85]; P = .003).

“and when a factor was present, the likelihood of incorporating it into the care plan also increased (aOR, 11.27 [95% CI, 2.27-55.85]; P = .003).”

Across groups, contextualized plans were associated with a significantly lower likelihood of a worsened red flag outcome (aOR, 0.67 [95% CI, 0.45-0.99]; P = .046).

“and a significantly lower likelihood of a worsened red flag outcome (aOR, 0.67 [95% CI, 0.45-0.99]; P = .046).”

When the red flag was the subject of an alert, it was less likely to have worsened 4 months later (aOR, 0.19 [95% CI, 0.05-0.73]; P = .02).

“In addition, when the red flag was the subject of an alert, it was less likely to have worsened 4 months later (aOR, 0.19 [95% CI, 0.05-0.73]; P = .02)”

When the red flag was the subject of an alert, it was not more likely to have been improved (aOR, 0.75 [95% CI, 0.30-1.89]; P = .55), beyond the effect of whether the care plan had been contextualized.

“but not more likely to have been improved (aOR, 0.75 [95% CI, 0.30-1.89]; P = .55), beyond the effect of whether the care plan had been contextualized.”

There was no effect of whether the recorder was visible or concealed on incorporation into care plans (aOR, 0.86 [95% CI, 0.28-2.66]; P = .79).

“There was no effect of whether the recorder was visible or concealed (aOR, 0.86 [95% CI, 0.28-2.66]; P = .79).”

Across groups, contextualized plans were associated with a significantly higher likelihood of an improved red flag outcome (aOR, 2.13 [95% CI, 1.38-3.28]; P = .001).

“However, across groups, contextualized plans were associated with a significantly higher likelihood of an improved red flag outcome (aOR, 2.13 [95% CI, 1,38-3.28]; P = .001) (Table 1 in eAppendix 6 in Supplement 2)”

Clinicians address contextual factors 60% to 70% of the time.

“Clinicians address them 60% to 70% of the time, leaving substantial room for improvement.6,7”

• cites [6] Weiner SJSA, Schwartz A, Sharma G,. Patient-centered decision making and health care outcomes: an observational study. Ann Intern Med. 2013; 158(8): 573- 579. doi: 10.7326/0003-4819-158-8-201304160-00001 23588745 sourceNot checked — no full text of the source
• cites [7] Weiner SJ, Schwartz A, Weaver F,. Contextual errors and failures in individualizing patient care: a multicenter study. Ann Intern Med. 2010; 153(2): 69- 75. doi: 10.7326/0003-4819-153-2-201007200-00002 20643988 sourceNot checked — no full text of the source

The finding that physicians pay less attention to contextual red flags presented to them by CDS than those presented directly by the patient points to a limitation of this technology.

“The finding that physicians pay less attention to contextual red flags presented to them by CDS than those presented directly by the patient points to a limitation of this technology.”

The study had 80% power to detect an absolute increase in contextualization rates from 50% to 75% with recordings from 192 intervention and 288 control patients.

“Based on the number of contextual red flags and factors we expected, we determined that to obtain 80% power to detect an absolute increase in contextualization rates when factors were present from 50% to 75%, we would need recordings from 192 intervention and 288 control patients (which also provides at least 80% power to detect the expected increase in probing).”

• cites [6] Weiner SJSA, Schwartz A, Sharma G,. Patient-centered decision making and health care outcomes: an observational study. Ann Intern Med. 2013; 158(8): 573- 579. doi: 10.7326/0003-4819-158-8-201304160-00001 23588745 sourceNot checked — no full text of the source
• cites [16] Weiner SJ, Schwartz A, Sharma G,. Patient-collected audio for performance assessment of the clinical encounter. Jt Comm J Qual Patient Saf. 2015; 41(6): 273- 278. doi: 10.1016/S1553-7250(15)41037-2 25990893 sourceNot checked — no full text of the source

Approximately 40% to 50% of patient presentations contain contextual factors.

“Although not all patient presentations contain contextual factors, approximately 40% to 50% do.6”

• cites [6] Weiner SJSA, Schwartz A, Sharma G,. Patient-centered decision making and health care outcomes: an observational study. Ann Intern Med. 2013; 158(8): 573- 579. doi: 10.7326/0003-4819-158-8-201304160-00001 23588745 sourceNot checked — no full text of the source

Because coders are blinded as to study group, the subjectivity of some outcomes should not skew the results.

“Because coders are blinded as to study group, this limitation should not skew the results.”

This is the first study to demonstrate that CDS tools built into the EHR can decrease contextual errors.

“Although it has been previously established that contextual errors are common, that they adversely affect patient outcomes, and that clinicians can learn to make fewer of them with feedback,8 this is the first study, to our knowledge, to demonstrate that CDS tools built into the EHR can decrease contextual errors.”

Loss of control of a chronic condition, missed appointments, and frequent emergency department visits are contextual red flags that improve when their underlying contextual factors are addressed.

“Loss of control of a chronic condition, missed appointments, and frequent emergency department visits are all examples of contextual red flags that improve when their underlying contextual factors are addressed.6”

• cites [6] Weiner SJSA, Schwartz A, Sharma G,. Patient-centered decision making and health care outcomes: an observational study. Ann Intern Med. 2013; 158(8): 573- 579. doi: 10.7326/0003-4819-158-8-201304160-00001 23588745 sourceNot checked — no full text of the source

The real effect of the intervention may be larger than what was found, because patients in both study groups completed the questionnaire, which may have primed control group patients to raise contextual red flags and factors.

“Finally, the real effect of the intervention may be larger than what was found, because we had patients in both study groups complete the questionnaire, which may have primed those in the control group to raise contextual red flags and factors.”

The lack of an effect of the visible recorder on clinician probing suggests that clinicians were not motivated to change their behavior just because they knew they were in a study.

“Furthermore, the lack of an effect of the visible recorder on clinician probing suggests that they were not motivated to change their behavior just because they knew they were in a study.”

Trained coders achieve approximately 85% interrater agreement about whether a care plan is contextualized.

“Trained coders achieve approximately 85% interrater agreement about whether a care plan is contextualized or not.15”

• cites [15] Weiner SJ, Kelly B, Ashley N,. Content coding for contextualization of care: evaluating physician performance at patient-centered decision making. Med Decis Making. 2014; 34(1): 97- 106. doi: 10.1177/0272989X13493146 23784847 sourceNot checked — no full text of the source