Exclusive: The Search for Reliable Health Information, Part V

The reliability of the current state of health information is extremely limited.

EDITOR’S NOTE: Dr. Joseph Nichols is producing a five-part series on healthcare data for ICD10monitor. This is the fifth installment in his exclusive series.

This is the last article of the five-part series on “the search for reliable health information.” This discussion about the impact of bias on information quality is brief, but probably the most important consideration toward reaching the desired goal of reliable and meaningful information that can be used to drive health policy and actions.

Knowledge should be derived independently, purposefully blind to the potential outcome of any critical analysis of accurate data. Knowledge never defines what needs to be done; it takes human judgment based on unbiased information to make that determination

Bias is arguably the hardest obstacle to overcome, and it has the greatest impact on information quality. It’s better to have no information, as opposed to information that’s intended to support a belief that may not be true. Biased information can easily cement a belief that is wrong and result in failure to act, or action that leads to the wrong outcome. It is human nature to cling to a belief, often with great passion, that may or may not be supported by solid evidence. Beliefs are comforting, and give us a sense of knowledge and control. We shun the unknown and replace it with belief. Once a belief is established, we look for information to confirm it, and reject information that doesn’t support it. It’s a tendency that is hard to avoid. Often, this bias is unconscious, but not infrequently, bias comes in the form of data or analytic manipulation intended to prove an agenda. It takes serious effort to mitigate this bias so that we can replace unsubstantiated beliefs with new knowledge that can improve decisions.

Commitment to unbiased information

  • Data and information governance require leadership that is committed to honest discovery, even if that information doesn’t support a preconceived notion favored by the organization. This usually means that governance over information must be given independence from the organization’s upper management. Information governance needs to be protected from risk, when information may prove uncomfortable or challenges longstanding beliefs. If, for example, objective, quality data analysis points to serious quality issues, the information should guide change, and not be set aside as incorrect or inconsequential. Undiscovered problems can’t be corrected.
  • Today’s news is replete with examples of manipulated data or analysis of data to drive home a specific agenda. In most instances, that manipulation results in a delayed or failed action that may be critical to health or safety.
  • Politics, greed, fear, pride, and desire for power are major drivers of bias on a regular basis. It requires persistent discipline to overcome these drivers.

Potential Solutions

  • Create independent information governance that reports directly to a board or other oversight entity without fear of reprisal from direct management. The federal government has attempted to use this type of model through entities such as the OMB (Office of Management and Budget) or the OIG (Office of the Inspector General), with variable success. The purpose of this independent oversight is to protect the population from wrong-minded decisions that run counter to the direction suggested by independent analysis of accurate data. Recent history has shown how challenging such analysis can be, in the face of strong political and power-driven forces, but it is critically necessary.
  • Maintain a discipline of driving out bias throughout each phase of the information process: from source data, aggregation, or categorization, to analysis and reporting. Bias that is introduced at any step can compromise knowledge needed to drive higher-level, publicly declared goals. Statistics must follow disciplined principles. Trends and differences should not be declared unless there is valid statistical support. There is a common statistical statement that “given two numbers, they will usually be different.” A variation across two or three data points in any direction does not indicate a trend. Depending on the size and variability of the underlying data, it can take a remarkable number of data points in any direction to reliably establish a trend. The risk lies in not adhering to a statistical definition and instead simply declaring a trend, if it supports an established belief or agenda.
  • Establish the long-range value of embracing information from reliable and properly analyzed sources toward improving actions that are consistent with overarching goals. Quality measures may reveal quality problems in the short term, but proper action can demonstrate remarkable improvement over time. Additionally, attempts to hide or manipulate facts for a short-term desirable result usually end up being discovered, and can result in an even more compromising situation.


Bias may be the ultimate stumbling block. There is a strong tendency to attempt to drive information to support our predetermined beliefs and agendas. Overcoming bias requires discipline, independence, authority, and ultimately, an agreement that unbiased information can improve healthcare value. That critically important process will not happen unless leadership drives home the principle that getting at the truth, even if it hurts, is a desirable goal.


Print Friendly, PDF & Email

You May Also Like

Leave a Reply

Please log in to your account to comment on this article.


Subscribe to receive our News, Insights, and Compliance Question of the Week articles delivered right to your inbox.

Resources You May Like

Trending News