Patient Satisfaction News

Assessing CAHPS Surveys, Patient Satisfaction with Machine Learning

Organizations can deploy their own mock CAHPS surveys and use machine learning to identify problem areas and at-risk patient populations to quell negative responses on the true CAHPS survey.

cahps survey

Source: Thinkstock

By Sara Heath

- A machine learning approach to understanding CAHPS survey responses may help health plans and organizations make targeted approaches to improving patient satisfaction, according to a recent report from Decision Point.

The report, “Impacting Perceptions of Healthcare Access & Satisfaction,” is a patient engagement playbook that guides insurance plans through using machine learning techniques that link certain parts  of CAHPS surveys with different health system’s functions.

High patient satisfaction scores do not solely depend upon clinicians executing their job functions well. High scores also require clinicians and health plans to understand how their jobs meet the unique needs of different patient populations.

“High CAHPS ratings starts with having a high-quality healthcare organization, with excellent access to care and meaningful processes in place to promote satisfaction,” the playbook noted. “Even with all this, however, CAHPS is so dependent on the demographic, utilization and disease profile of the population, that having a high quality organization is simply not enough.”

Ultimately, improving CAHPS scores boils down to four central areas:

  • Understanding the profiles of individuals that are negative responders;
  • Anticipating every individual’s survey response even if you do not know if they will be selected for the CAHPS survey;
  • Understanding the potential reason why a person will respond in a specific manner; and,
  • Determining a meaningful action plan to communicate with individuals and change behavior and potential survey responses over time.

READ MORE: How Hospitals Can Raise Patient Satisfaction, CAHPS Scores

“Because health plans are judged by consumers, and in some cases, compensated by CMS on the results of the CAHPS survey, it’s fiercely important for plans to develop a strategy to improve perceptions of the plan and its network of providers across the entire member base,” Saeed Aminzadeh, founder and CEO of Decision Point Healthcare Solutions, said in a statement.

Receiving high CAHPS scores is not always easy. The survey uses only a portion of randomly-selected health plan members. Health plans do not know who will be surveyed or who was surveyed in the past, making it hard for plans to pinpoint problem areas driving certain results.

Health plans that conduct broad outreach might expend a lot of resources for a very limited payout. Conversely, plans that attempt to target a certain patient population might not be reaching out to the correct people who will respond to a CAHPS survey.

“While it may be impossible for the plan to know which individuals responded positively or negatively to CAHPS and which members will be surveyed in the future, plans can beat the system by issuing their own ‘mock CAHPS’ survey using the same (or similar) survey questions,” the report recommended. “By doing this, the plan can link member responses to their clinical, utilization, and consumer profile and start profiling negative CAHPS responders.”

Machine learning can help health plans understand these mock CAHPS surveys and identify specific patient populations. When health plans determine perceived care quality and can see the demographics expressing certain concerns, they can tailor their outreach efforts to yield optimal results.

READ MORE: How Communication, Relationships Impact Patient Satisfaction

“Plans can use machine learning techniques to extrapolate the results of the survey to all members in the plan,” the report explained. “In short, machine learning helps identify members that exhibit similar behavior to members responding negatively to the CAHPS survey”

The first step to doing this is classifying patient responses to the mock CAHPS survey into categories: access to care issues, provider issues, plan issues. This will allow the health plan to work with the patient on her specific issues with the healthcare system.

Next, health plans must understand patient populations most “at risk” for reporting negative experiences on a CAHPS survey. These risk factors include:

  • Tenure with health plan. Longer plan members tend to have more leniency when completing CAHPS surveys.
  • Patients under the age of 65 tend to respond more negatively to CAHPS surveys, despite their tenure with the plan.
  • Patients with no or a limited history with a primary care provider tend to have more negative CAHPS responses. Limited PCP engagement might be tied to a dislike of their PCP or an inability to access and schedule appointments with their PCP.

The final leg of this approach is addressing specific CAHPS gripes. However, it is not feasible to address each individual issue with each individual patient, the report acknowledged. Instead, organizations should group their patient outreach goals with CAHPS improvement goals.

For example, a health plan can combine an appointment scheduling campaign with additional information that could close gaps in care access complaints.

READ MORE: Do Patient Satisfaction Scores Truly Portray Quality Care?

Healthcare organizations can also identify areas where patient complaints about their perceptions of care are deeply rooted in a health system. Patient complaints about long wait times might tie back to a single provider who has longer wait times. This issue can be addressed with that provider.

Reaching out to patients using both broad campaigns and targeted interventions may reduce disenrollment rates, the report noted. Additionally, it may improve the rate at which patients undergo preventive screenings, which in turn can reduce high cost burden on the health plan.

CAHPS scores have serious implications for health plans. Not only do these scores determine some of the reimbursements that plans receive from CMS, but CAHPS scores also inform health plan quality star ratings. Those quality star ratings were designed for consumers to make decisions about their health plan choices, and a poor star rating can determine whether or not a consumer purchases the health plan.

When a health plan can use machine learning to better predict who is answering CAHPS survey questions in a certain manner, they can pinpoint where their problem areas are. This will in turn improve the likelihood that a health plan can improve their patient satisfaction scores and retain more beneficiaries.


Sign up for our free newsletter:

Our privacy policy

no, thanks

Continue to site...