by Senem Guney, Zach Childrers, and Thomas H. Lee

“Happy families are all alike; every unhappy family is unhappy in its own way.”

With that opening line in Anna Karenina, Tolstoy began his exploration of the marriages and relationships in nineteenth-century Russia. As remote as that era may seem today, Tolstoy’s initial observation remains relevant in other contexts where success cannot be guaranteed, and failure can occur in multiple ways. This “Anna Karenina Principle” has been described in evolutionary biologymarketing research, and efforts to eliminate gender disparity in science and technology. The Anna Karenina Principle applies to health care, too, and has direct implications for how health care organizations should think about the patient experience.

Our analysis of patients’ own words about their care experiences reveals that certain positive themes should characterize every patient’s care. And, as the Anna Karenina Principle would suggest, a wide variety of missteps can unsettle patients and compromise their trust in their care. The small number of common themes that define positive patient experiences and the much larger number of ways in which care can disappoint patients suggest that organizations should employ the high reliability principles used to improve patient safety to improve patient experiences.

When the safety movement began about two decades ago in health care, clinicians started learning how to prevent medical errors that cause serious injury or death. One key step in preventing “Never Events” (events that should never happen such as performing surgery on the wrong site or operating on the wrong patient) was reliably performing “Always Events” — e.g., by having the whole team pause before any procedure and go through the World Health Organization Surgical Safety Checklist. An additional major theme that emerged in patient safety was building a learning organizational culture in which errors were reliably detected, their causes were analyzed, and steps were taken to prevent them in the future.

These complementary approaches — always performing certain desired actions, combined with learning from adverse events when they occur — should also be part of any organization’s efforts to improve the patient experience. Health care organizations should try to give every person care characterized by the themes that dominate positive patient experiences. And they should work relentlessly to learn from negative experiences. In short, they should apply the Anna Karenina Principle and work both to create good patient experiences while also avoiding bad ones.

The relevance of the Anna Karenina Principle to patient experience can be seen in phrase-by-phrase analysis of patient comments that we have been conducting with an artificial intelligence (AI) platform that uses linguistic rules of sentence structure and machine learning algorithms. Here’s one example: We analyzed 175,334 patient comments collected through surveys performed by our organization (Press Ganey) at one major teaching hospital between November 2018 and March 2020. We focused on surgical units, because patients’ needs as they recover from procedures are more homogeneous than in the overall patient population. Our analysis first identified positive and negative sentiments and then the themes and subthemes running through the positive and negative sentiments to show the underlying main reasons the patients felt that way.

The first striking finding from these analyses is the consistency among the positive comments — the characteristics of care noted by what Tolstoy might have called happy patients. On every floor, the most common positive theme was that patients felt that they had been treated with courtesy and respect. This pattern in patient experience comments is found consistently across the country. Other analyses of Press Ganey data demonstrate that patients in every setting (e.g., inpatient, outpatient, emergency department) value three closely related themes: empathy, coordination of their care, and good communication.

The second striking finding was the marked variation among the negative comments — the characteristics of care noted by what Tolstoy might have called unhappy patients. For example, the themes from negative comments varied markedly across the five surgical units, each of which had a different major problem area. For some units, the most frequently noted source of unhappiness was long waits for assistance. In others, the most common issue was noise. In still others, patients complained about chaos with the discharge process.

These data demonstrate that preventing negative experiences requires the same kind of vigilance needed to prevent the vast range of potential safety problems, which vary from pressure ulcers to falls to hospital-acquired infections. Variation abounds. One specific issue may emerge as the most common problem on one patient care unit, while another issue is likely to emerge on another unit — and next week, the problems might be reversed. A bad experience — e.g., a patient feeling vulnerable to catching Covid-19 because the bathrooms are not clean — can ruin a good experience created through technically excellent and empathic care.

The nature of these challenges calls for complementary approaches to data collection and analysis. Structured questions in which patients are asked to rate their experiences on Likert-type scales (e.g., using 1-5 scales, with 5 being extremely positive and 1 extremely negative) are efficient for assessing whether positive themes are occurring with reliability. Simply put, a good way of checking to see if patients were treated with courtesy and respect is asking them if they felt that they were treated with courtesy and respect.

But AI-enabled analysis of narrative data can capture insights and nuances that go beyond the information available from survey questions by extracting insights from thousands of comments — and is thus particularly valuable for understanding what makes unhappy patients unhappy. For example, patients’ check-box answers to survey questions can identify cleanliness as a problem on a patient care unit. Comments, however, can pinpoint personal hygiene (e.g., gowns and beddings soiled from blood or other bodily fluids) or litter or staff behaviors as specific issues.

These two types of data collection and analysis support caregivers in their two types of efforts to earn patients’ trust. Clinicians cannot always provide cures or even complete relief of suffering, but it is well within their control to be reliable on themes such as courtesy, respect, empathy, coordination, and communication. Ensuring that these positive themes characterize every patient’s care requires the establishment and reinforcement of social norms for caregivers. That work begins at the top for every organization — i.e., leaders must show how important these values are to them, and managers must ensure that they are standards of care, not mere guidelines or recommendations. At many organizations, “disrespect events” in which patients are not treated with dignity, are reviewed in the same daily meetings as safety events and near misses.

This form of “high reliability” should be complemented by systematic efforts to prevent the wide range of lapses that can be unsettling to patients — as opposed to just abhorring them. When a patient is given contradictory information from different caregivers (one says they are going home tomorrow, and another says they are going to a rehabilitation facility), that is not something to be shrugged off. An error has occurred; it may be no particular individual’s fault, but it is a system failure — and one that understandably shakes patients’ confidence in their care. And it should be detected and analyzed, and the response should be akin to the “error trapping” approaches developed in patient safety programs.

The work of preventing the many possible causes of failure from the perspective of patients is enormous and can be consuming. And it is worth noting that most of the 800-plus pages of Anna Karenina are devoted to family unhappiness. But the work of preventing failures should not distract health care organizations from the importance of ensuring the reliability of positive themes. To make care excellent, we must be highly reliable in developing cultural norms and preventing failures so that every patient can be like a member of one of Tolstoy’s “happy families.”


Originally published in Harvard Business Review.