Data mining nursing care plans of end-of-life patients: a study to improve healthcare decision making

Abstract

Purpose: To reveal hidden patterns and knowledge present in nursing care information documented with standardized nursing terminologies on end-of-life (EOL) hospitalized patients.

Method: 596 episodes of care that included pain as a problem on a patient's care plan were examined using statistical and data mining tools. The data were extracted from the Hands-On Automated Nursing Data System database of nursing care plan episodes (n = 40,747) coded with NANDA-I, Nursing Outcomes Classification, and Nursing Intervention Classification (NNN) terminologies. System episode data (episode = care plans updated at every hand-off on a patient while staying on a hospital unit) had been previously gathered in eight units located in four different healthcare facilities (total episodes = 40,747; EOL episodes = 1,425) over 2 years and anonymized prior to this analyses.

Results: Results show multiple discoveries, including EOL patients with hospital stays (<72 hr) are less likely (p < .005) to meet the pain relief goals compared with EOL patients with longer hospital stays.

Conclusions: The study demonstrates some major benefits of systematically integrating NNN into electronic health records.

Figure 1. An abstract diagram showing different entities and their inter-dependencies in the HANDS database

Each box shows a unique feature being stored in the database for each patient and/or care provider. Copyright 2012 HANDS Team, used by permission.

Figure 2. Percentage of EOL patients (all hospitals) meeting the expected pain level ratings at discharge

More than half of the patients do worse than expected.(Χ²(1)=7.49, p-value: 0.0006).

Figure 3. Percentage of patients meeting their pain related NOC expected outcome ratings by their length of stay

(short < 72 hours, medium 72 to 199 hours, and long >= 200 hours) across all four hospitals, p-value: < .001.

Figure 4. The impact of different combinations of NIC interventions on the pain related NOC outcome ratings

Shown are the numbers of patients who received the different combinations of NIC interventions by hospital andwhether they met or failed to meet their expected pain level rating (at discharge).In this example, adding Positioning (P) to Medication Management (MM) and Pain Management (PM) resulted in a statistically significant greater likelihood of meeting the expected pain-level rating at discharge or death.

Figure 4. The impact of different combinations of NIC interventions on the pain related NOC outcome ratings

Shown are the numbers of patients who received the different combinations of NIC interventions by hospital andwhether they met or failed to meet their expected pain level rating (at discharge).In this example, adding Positioning (P) to Medication Management (MM) and Pain Management (PM) resulted in a statistically significant greater likelihood of meeting the expected pain-level rating at discharge or death.

Figure 5. Predicting patient’s discharge pain rating from rating at 24 hours post admission

Using the patient’s pain rating at 24 hours post admission this table shows the distribution of expected pain-level ratings at discharge or death, and the percentages of patients meeting or failing to meet their expected ratings. The highlighted cell shows a patient with medium pain at the 24-hour point who is expected to have no pain by the time of discharge. 36% of patients fall into this category, but of those only 1 in 6 will meet the pain free goal. Copyright 2012 HANDS Team, used by permission.

Figure 6. User Interface of an example decision-making assistant based on knowledge discovery using data mining of Nursing Care Plans

Examples shows that system has identified a patient with pain-trends that match with the historic data, and tags the patient data records with a red arrow and based on the data mining outcomes suggests interventions that are known to work well for such patients. Copyright 2012 HANDS Team, used by permission.