A systematic review on the effectiveness and impact of clinical decision support systems for breathlessness

A total of 4294 records were screened, 127 underwent full-text review and 37 studies were included (Fig. 1). Many studies were excluded as they report only retrospective validation22 or was purely for patient use and did not inform clinicians23. Majority of the studies (n = 30) included were RCTs. Study duration ranged from 2 weeks24 to 5 years25. About 55% of the studies (n = 20) were conducted in the United States followed by the Netherlands (n = 6)26,27,28,29,30,31 and three each in Canada32,33,34, South Korea35,36,37, and United Kingdom38,39,40. All were in high-income countries and mostly in urban settings (Supplementary Table 2).

Fig. 1figure 1

Most studies were conducted only in primary care or had a primary care arm (n = 20), followed by 13 in the ED and four in a mixed setting. CDSS interventions described in the studies varied from non-electronic support to alerts, expert systems, and artificial intelligence-based CDSS. About half (n = 17) referred to either hospital, national or international guidelines.

Only one study explicitly addressed breathlessness. The majority were for asthma (n = 19) or heart failure (n = 9). Most focused on a single disease with seven addressing more than one disease.

Outcomes from CDSS for multimorbidity

Majority of the studies (6/8) were cluster RCTs and conducted in primary care. Some CDSS focused on a combination of diseases such as asthma and angina, others on symptoms such as breathlessness and chest pain, and on diagnostic test support such as spirometry. (Supplementary Tables 2 and 3).

Studies reported mixed results. In studies with no significant difference in the CDSS arm compared to control arm, a low level of software use (median active interactions with the CDSS = 0)40 and large variations between GPs’ prescribing behavior which underpowered studies28 was reported. Process outcomes (guideline compliance) were reported to significantly improve in a cardiovascular CDSS study41, but an earlier study reported otherwise42.

In the emergency department (ED), a multi-center RCT by Kline et al.43,44 reported no significant difference in median length of stay in the ED, but median length of stay in the hospital was significantly lower in the intervention group (7.7 h [IQR 4.0–27.3] vs 8.9 h [IQR 4.8–29.6], p = 0.046).

Outcomes from asthma-specific CDSS

Eighteen studies found were focused on asthma. There was an equal mix of those conducted in children and adults.

In children, CDSS use for diagnosis was found to improve the proportion diagnosed with asthma45 but reported mixed impact on exacerbations46,47, symptom days48 and influenza vaccination rates49 post-implementation. However, a pilot study50 reported improvement in patient-reported outcomes such as missing days from work and quality of life. A few studies focused on management, especially measuring adherence to guideline recommendations. Some studies showed improved adherence to guidelines on spirometry51, AAPs51, peak-flow measurements29, and oxygen saturation measurement24.

In adults, a diagnostic study37 reported a CDSS sensitivity of 85.2% and specificity of only 25% in ascertaining a diagnosis of asthma. As in children, several studies have reported improved guideline adherence in the CDSS arm. Studies reported improvement in peak-flow measurements29, inhaled corticosteroids (ICS) prescription32,36,52, and use of AAPs52. They also reported improved symptoms, Asthma Quality of Life Questionnaire (AQLQ) score, forced expiratory volume one second (FEV-1) value, airway hyperresponsiveness52, and reduced exacerbations39. The benefits on hospitalization were mixed 33,39.

Outcomes from COPD-specific CDSS

Two studies26,53 focused on patients with COPD. Both reported improvements in patient-reported outcomes upon CDSS implementation.

Outcomes from heart failure-specific CDSS

Eight CDSS focused on heart failure patients, for various uses ranging from diagnosis to management as well as being linked to telemonitoring systems. A diagnostic study using an artificial intelligence-based CDSS35 found a significantly higher accuracy of the CDSS compared to usual care (97.9% vs 76.3%). Studies also showed benefits on improving cardiac rehabilitation31 and management guideline compliance54.

One RCT54 among patients with heart failure reported significantly lower mortality in the CDSS compared to control group (2% vs 14%), although no significant differences in hospitalization and ED visits were found between groups. Another RCT55 reported a paradoxical increase in all-cause hospitalizations which the investigators attributed to similar levels of guideline adherence between intervention and control groups (33% vs 30%, p = 0.4).

Two studies30,38 evaluated integrating CDSS with telemonitoring. Both reported no significant difference in outcomes, but one did show a significantly lower number of heart failure-related outpatient clinic visits (median 2 vs 4, p = 0.02)30.

Health economic outcomes

The cost of managing breathlessness can arise from both direct costs (healthcare use, medication costs) and indirect costs (productivity loss)(Supplementary Table 4). The majority of studies (seven) reporting this outcome focused on asthma, with two on heart disease and one on breathlessness in the ED.

For asthma, studies reported lower number of missed days at work and school in the CDSS group than control group48,50, with mixed results on the impact on the number of health visits29,39. One cost-benefit analysis48 reported that by year 1, the savings (USD337/child) from the reduction in ED visits from a CDSS intervention were sufficient to provide a cost-benefit. Montecarlo simulation suggests a 97% chance of this cost saving. Other studies reported mixed results, some reported no difference46, while others a higher cost24,56 in the intervention group with wide variations in cost.

Two studies on patients with heart disease reported a cost-effectiveness analysis. One30 reported that the addition of telemonitoring did not lead to greater cost-effectiveness. Another42 reported that while total health charges were lower for the intervention compared to control group, this was not statistically significant due to the wide variation in outpatient and inpatient costs reported in the study.

One multimorbid CDSS study conducted in the ED43,44 reported significantly lower median costs for medical care within 30 days in the intervention compared to the control group with small lifetime gains in Quality Adjusted Life Years.

Effect on physicians

Various studies assessed the impact on physicians differently (Supplementary Table 5). Studies reported that CDSS use improved knowledge33,36 and were found to be useful for daily practice33. Low rates of CDSS use were however reported in many studies41,55. High inter-doctor variation in prescribing behavior was found28 and providers were reported to be more compliant to reminders barring prescription of certain drugs rather than those that recommend certain medication.

Qualitative studies39 report that management recommendations and reminders were popular with the users and were deemed to provide relevant clinical advice. Physicians were reported to differentially use CDSS and were more likely to use it in out-of-control patients32. Physicians’ attitude towards guidelines underlying clinical recommendations were mixed. They ranged from viewing guidelines as providing helpful information but not helpful in making decisions for individual patients42,56 to CDSS recommendations not meeting patient needs25,41,55. Studies suggest that GPs have differing needs and that GPs handling more complex conditions may be more amenable to using CDSS40,41.

Despite this, studies also show that CDSS can save time41,46 and that providers had the impression that CDSS allowed them to accomplish more work than would otherwise be possible41. Although visits tend to last longer in the CDSS group, a CDSS can still be used in consultations lasting <10 min39.

Effect on patients

Studies evaluated patient satisfaction through various measures. In patients with heart disease, Tierney et al.42 found no significant difference in quality of life, medication compliance, and satisfaction with care between the intervention and control groups. Similarly, Breathett et al.’s study57 in heart failure patients found that while median patient satisfaction was higher in the intervention group, this was not statistically significant (p = 0.08). Even so, the study reported that patients particularly felt providers offered better explanations of their care in the intervention than control arm (83.7%, vs 55.8%, p = 0.01). They reported that patients in the intervention group perceived significantly better descriptions of medication side effects than the control group (61.1% vs 26.7%, p = 0.01).

Subramanian et al.55 study among heart failure patients also reported similar results. At six months, patients in the intervention group were more satisfied with their physicians (p = 0.02) and primary care visit (p = 0.02). However, at 12 months, only satisfaction with the most recent primary care visit remained statistically significant (p = 0.01).

Slok et al.26 in a study on COPD patients utilizing the Patient Assessment of Chronic Illness Care (PACIC; a measurement of perceived quality of care) score reported a 0.32 point (95% CI 0.14 to 0.50) improvement in the intervention group (scores range from 1–5).

In contrast, asthma patients in three studies assessing patient satisfaction reported no significant difference in activation score50, partnership problems58, or satisfaction in general56. The use of a patient-facing CDSS kiosk was reported to not improve partnership with providers58. The authors report that providers’ inattention to parents’ concerns communicated via the kiosk may explain a trend toward worsening partnerships noted in the adjusted results. Improvements in information sharing scores only occurred in the subset of patients whose kiosk output was acted upon by providers. The use of a CDSS for breathlessness in the ED43,44 was also reported to not improve patient satisfaction survey score (p = 0.148).

Unintended consequences of CDSS use on clinical practice

None of the studies reported serious adverse events resulting in death or breach of confidentiality in both intervention and control groups46. However, there were differential impacts of CDSS implementation on patients depending on their location51, age29,33, and regularity in visiting a health facility33 (Supplementary Table 6) This relates to findings from other studies29,31 which discussed how system-wide factors are also possible limitations to CDSS effectiveness. Furthermore, studies have also reported an increase in unscheduled visits52, and both higher dosing52 and underdosing54 of patients. For those with a patient-facing component, lowering engagement was also reported 34.

Risk of bias

The majority of RCTs were classified as low risk of bias (21/30 studies); eight had some concerns and one had a high risk of bias. For observational studies, three were assessed to have a serious risk of bias, one moderate risk, and one low risk. For the two diagnostic studies, one was assessed to be high risk and another low risk. Further details are in Appendix 2.

留言 (0)

沒有登入
gif